The End-to-End Walkthrough

The process begins with diagnosis, the step organizations most often skip and the one Rumelt (2011) identifies as the irreducible core of strategy, in which the product leader states plainly the central challenge the product faces rather than the goals it aspires to. Diagnosis draws on the structural analysis of the Five Forces to characterize the profit pool and the competitive dynamics, on the assessment of moats to determine what defensibility the product holds and lacks, and on an honest reckoning with the AI transition to identify whether the product’s value rests on capability that is commoditizing. A diagnosis is complete when it names the one or two conditions that most determine the product’s future, since a diagnosis that lists everything has diagnosed nothing.

The process continues with the guiding policy, the overall approach chosen to address the diagnosed challenge, which is where the reframing frameworks earn their place, since the guiding policy frequently embodies a blue-ocean reconception of the market, a decision about which network effect or moat to build, or a choice of where on the AI-native to AI-enabled spectrum the product will sit. The guiding policy is not a goal but a method, expressing how the product will win rather than what it hopes to achieve, and its quality is tested by whether it rules things out, since a guiding policy compatible with every possible action is not guiding anything.

The process then specifies coherent action, the set of concrete, mutually reinforcing moves that enact the guiding policy, which is where prioritization, portfolio allocation, business model, pricing, go-to-market, and organizational design become the instruments of execution rather than separate concerns. Coherence is the demanding criterion here, since the actions must reinforce one another rather than merely coexist, and a strategy whose pricing fights its go-to-market or whose team structure contradicts its guiding policy will be ground down by its own internal friction. The walkthrough closes with the explicit articulation of the assumptions on which the strategy depends and the leading indicators that will reveal whether it is working, drawing on the uncertainty and metrics modules so that the strategy is built to be tested rather than merely asserted.

Templates

The following template structures the strategy document itself, designed to fit the one-page-narrative discipline while accommodating the depth a longer memo requires.

The second template supports the diagnosis step by structuring the situational analysis the document compresses, and it is completed before the document is written rather than after.

Constructing the Strategic Narrative

The strategic narrative is the prose that carries the strategy into the minds of the people who must act on it, and its construction follows the storytelling discipline of the leadership modules rather than the structure of the analytical worksheet. An effective strategic narrative opens from a situation the audience recognizes, introduces the tension that the diagnosis has identified, and arrives at the guiding policy as the natural resolution of that tension, so that the audience experiences the conclusion as discovered rather than asserted. The narrative anchors the abstract strategy in a concrete scene, frequently the experience of a specific customer in the future the strategy aims to create, since this is what converts the strategy from a proposition the audience evaluates into a future the audience can inhabit. The discipline of the narrative is intellectual honesty, since it must acknowledge the genuine uncertainties and the strongest objections and address them, which is what earns the trust of the skeptical executives whose conviction the strategy requires, and it must be compressible to the single page that makes it portable through the organization without the author present.

A Decision Tree for the Core Strategic Choices

The following decision tree organizes the sequence of choices a product strategy must make, ordering them so that each decision constrains the next, which is how a coherent strategy is built rather than assembled.

The tree is not a substitute for judgment but a scaffold for it, ensuring that the strategy addresses defensibility before growth, fit before scale, and the AI condition before everything else, which is the ordering the series has argued for throughout.

Closing the Series

The synthesis with which this series concludes is that product strategy is neither the application of a single favored framework nor the accumulation of many, but the disciplined integration of diagnosis, guiding policy, and coherent action, informed by the frameworks and grounded in the specific situation the product faces. The AI transition does not displace this discipline but intensifies it, since the commoditization of capability has raised the premium on genuine defensibility, the migration of work to agents has unsettled business models and category definitions, and the acceleration of change has made strategy under uncertainty the normal condition rather than the exception. The agenda for the product leader is to treat strategy as a document that is reasoned through rather than asserted, built on an honest diagnosis, expressed as a guiding policy that rules things out, enacted through coherent and mutually reinforcing actions, carried by a narrative that earns conviction, and held accountable to the assumptions and leading indicators that will reveal whether it is working. A strategy designed this way will sometimes be wrong, since the future is genuinely uncertain, but it will be wrong in ways the organization can detect and correct, which is the most that strategy can honestly promise and considerably more than most organizations achieve.

References

Lafley, A. G., & Martin, R. L. (2013). Playing to win: How strategy really works. Harvard Business Review Press.

Rumelt, R. P. (2011). Good strategy / bad strategy: The difference and why it matters. Crown Business.

Watkins, M. D. (2007). Demystifying strategy: The what, who, how, and why. Harvard Business Review. https://hbr.org/2007/09/demystifying-strategy-the-what

Netflix and the Willingness to Cannibalize

Netflix illustrates the precise discipline whose absence destroyed Kodak, namely the willingness to cannibalize a profitable business before a competitor or a technology shift does it for you. Netflix built a successful DVD-by-mail business and then deliberately undermined it by investing in streaming, a transition that threatened its own economics and its existing customer relationships, and it later compounded the move by investing in original content that risked its relationships with the studios that supplied its catalog. The strategic mechanism is the deliberate self-disruption that the portfolio module identified as the antidote to organizational gravity, and Netflix’s distinction is that it made the transformational bet while the core was still healthy rather than waiting until decline forced the move, which is the timing that determines whether self-disruption is a strategy or a desperate reaction. In the AI transition, Netflix’s lesson is the most directly applicable of any here, since the incumbents who will endure are those willing to cannibalize their profitable per-seat and human-operated businesses before agentic alternatives do.

Amazon and the Compounding Flywheel

Amazon illustrates the power of designing a business as a self-reinforcing system rather than a linear value chain, the growth-loop principle realized at the scale of an entire company. Amazon’s retail flywheel, in which lower prices drew more customers, who drew more sellers, whose competition lowered prices further, was a deliberately constructed compounding loop, and the company extended the same systemic thinking to Amazon Web Services, which emerged from the discipline of building internal capabilities as externalizable services that the Bezos API mandate had earlier enforced. The strategic mechanism is the construction of compounding systems whose advantage grows with scale, combined with the working-backwards discipline of reasoning from customer value, and Amazon’s durability comes from operating several reinforcing loops rather than depending on any single product. In the AI transition, Amazon’s position rests partly on the same systemic logic, since its infrastructure business benefits as the demand for compute compounds, illustrated by the scale of its multi-year compute partnerships with leading model developers (Amazon, 2025).

Stripe and the Developer as the Customer

Stripe illustrates the strategic power of identifying an underserved customer and serving them with uncompromising focus, in this case the developer who needed to accept payments without enduring the complexity that incumbents imposed. Stripe’s mechanism was to treat the application programming interface as the product and the developer as the customer, reducing the integration of payments from a months-long ordeal to a few lines of code, which is the API-as-product principle of the ecosystem module executed with exceptional discipline. The durable advantage Stripe built was the combination of developer trust, accumulated integration switching costs, and an expanding platform of financial services that deepened as customers grew, and its continued expansion into new financial infrastructure reflects the same focus on building tools that developers reach for first (IPE Newsletter, 2025). The lesson is that a relentless focus on a specific customer’s genuine difficulty, served better than anyone else serves it, builds a position that broad but shallow competitors cannot dislodge.

Figma and Collaboration as the Wedge

Figma illustrates the reframing of a market through a structural insight about how work actually happens, the blue-ocean and network-effect principles combined. Figma entered a design-tools market dominated by powerful incumbents not by building a better version of their single-player desktop software but by reconceiving design as a collaborative, browser-based, multiplayer activity, which created a network effect among designers, developers, and stakeholders who could now work in the same file. The strategic mechanism was to compete on a dimension the incumbents had structurally neglected, collaboration, and to convert it into a network effect that made Figma more valuable as more of a team adopted it, which is why the position proved so durable that a proposed twenty-billion-dollar acquisition by Adobe was abandoned under regulatory pressure in 2023, after which Figma went public independently in 2025 at a valuation that affirmed the strength of the position it had built (Figma, 2023; Figma, 2025). In the AI transition, Figma’s collaborative network is precisely the form of defensibility that AI does not erode, since the people in the network cannot be copied by a model.

Nvidia and the Ecosystem Moat

Nvidia illustrates the construction of an ecosystem lock-in so deep that even superior alternatives struggle against it, the switching-cost and ecosystem principles realized over nearly two decades. Nvidia’s strategic mechanism was to invest, beginning in 2006, in CUDA, a software platform that made its hardware programmable for general computation, and to cultivate a developer ecosystem that grew to over four and a half million developers whose accumulated code, libraries, and expertise are specific to CUDA (Nvidia, 2024). The durable advantage is that this ecosystem constitutes a switching cost borne not by Nvidia but by the entire community that has built upon CUDA, which is why Nvidia’s dominance in AI computation rests as much on the software ecosystem as on the hardware, and why competitors offering competitive chips contend with a barrier that hardware alone cannot overcome. The lesson is that the most durable moats are often built in the layer adjacent to the obvious product, in this case software surrounding hardware, and accumulated patiently over a horizon most companies will not sustain.

OpenAI and the Distribution of a New Capability

OpenAI illustrates the strategic power of pairing a genuine capability breakthrough with a distribution mechanism that converts the breakthrough into a market position before competitors can respond. OpenAI’s research produced capable models, but its strategic mechanism was the decision to package that capability into ChatGPT, a product that made the capability immediately accessible to an enormous audience, which converted a research advantage into a distribution and brand advantage of the kind that the AI strategy module identified as durable when capability itself is commoditizing. The continued investment in a developer platform that lets others build on the capability extends the position from a product into an ecosystem (OpenAI, 2025). The lesson, and the appropriate one with which to close a survey of winners, is that a breakthrough capability is necessary but not sufficient, and that the durable advantage comes from the distribution, the brand, and the ecosystem built around the capability rather than from the capability alone, which is the central strategic truth of the AI era that this entire series has developed.

The synthesis for the product leader is that winning strategies are instances of transferable mechanisms rather than inimitable visions, and that the mechanisms recur: the willingness to cannibalize the core that Netflix demonstrated, the compounding systems that Amazon built, the focused service of an underserved customer that Stripe perfected, the market reframing through a structural insight that Figma executed, the patiently constructed ecosystem moat that Nvidia accumulated, and the distribution of a new capability into a durable position that OpenAI achieved. The agenda is to read these not as stories to admire but as mechanisms to apply, asking which of them the organization’s own situation calls for, since the companies that endure are those that recognized the structural move their moment required and committed to it before the window closed.

References

Amazon. (2025). AWS and OpenAI announce multi-year strategic partnership. https://www.aboutamazon.com/news/aws/aws-open-ai-workloads-compute-infrastructure

Figma. (2023). Figma and Adobe are abandoning our proposed merger. https://www.figma.com/blog/figma-adobe-abandon-proposed-merger/

Figma. (2025). Figma announces pricing of initial public offering. https://www.figma.com/blog/ipo-pricing/

Nvidia. (2024). CUDA platform for accelerated computing. https://developer.nvidia.com/cuda

OpenAI. (2025). OpenAI for developers in 2025. https://developers.openai.com/blog/openai-for-developers-2025

Kodak and the Refusal to Cannibalize

The Kodak case is the most cited because it is the most damning, since Kodak did not fail to see the future but invented it and then declined to pursue it. A Kodak engineer, Steve Sasson, built the first digital camera in 1975, and the company’s internal study in 1981 correctly forecast that it had roughly a decade to prepare for the transition from film to digital, which means Kodak possessed both the technology and an accurate prediction of its own disruption (Killer Innovations, 2023). The mechanism of failure was not blindness but the refusal to cannibalize a profitable core, since digital photography threatened the film business that generated Kodak’s margins, and the organization’s incentives, structure, and identity were all bound to film, which made the rational pursuit of digital feel like the irrational destruction of the company’s most valuable asset. The lesson is that the most dangerous disruptions are those a company can see clearly, because the clarity does not overcome the organizational gravity that protects the profitable present against the uncertain future, which is precisely the dynamic the portfolio and prioritization modules earlier identified as the structural reason organizations starve their own transformational bets.

Nokia, BlackBerry, and the Misreading of the Basis of Competition

Nokia and BlackBerry failed through a related but distinct mechanism, namely the misreading of what customers had come to value once the basis of competition shifted. Nokia dominated mobile phones through hardware excellence and operational scale, and when the smartphone era arrived it held on to its Symbian operating system too long and moved to a modern software platform too late, having misjudged that the competition had shifted from hardware to software ecosystems (Medium, 2023). BlackBerry, whose physical keyboard and secure messaging had made it indispensable to professionals, dismissed the touchscreen as a fad unsuited to serious users, failing to recognize that the iPhone had redefined what a phone was from a communication device into a pocket computer whose value lay in its applications. Both companies were defeated not by a better version of the product they made but by a redefinition of the product category, which is the substitute threat of the Five Forces module realized, and the lesson is that incumbents are most vulnerable precisely when they evaluate new entrants against the old basis of competition, judging the iPhone a poor phone while it was busy becoming a different and superior thing.

Yahoo and the Cost of Strategic Indecision

Yahoo’s failure is the most diffuse and in some ways the most cautionary, because it was not the failure to see a single disruption but the chronic inability to decide what the company was. Yahoo possessed enormous audience, talent, and opportunity, and it dissipated these assets through a sustained indecision about whether it was a media company or a technology company, which produced a strategy that was perpetually hedged and never committed. The mechanism of failure here is the absence of the clear diagnosis and guiding policy that Rumelt identifies as the content of real strategy, since a company that cannot say what it is cannot allocate coherently, and Yahoo’s scattered acquisitions and reversals reflected the deeper failure to make the hard choice that strategy requires. The lesson is that indecision is itself a decision, and frequently the worst one, since a company that refuses to commit to a coherent identity surrenders the focus that allows resources to compound, and dissipates through diffusion advantages that a committed competitor concentrates.

WeWork and the Confusion of Narrative with Fundamentals

WeWork’s collapse is the most recent and the most instructive about a specifically contemporary failure mode, namely the substitution of narrative for fundamentals sustained until the moment of public scrutiny. WeWork reached a private valuation of forty-seven billion dollars on a story that it was a technology company transforming the future of work, when its underlying business of leasing real estate long and renting it short lost money on a model that depended on perpetual growth to obscure the losses (iDeals, 2023). When the company filed to go public in August 2019, the prospectus exposed the gap between the narrative and the fundamentals to institutional investors who, unlike the private backers who had funded the story, were evaluating risk rather than buying a vision, and the valuation collapsed from forty-seven billion to under ten within weeks, the offering was withdrawn, and the company eventually filed for bankruptcy in 2023. The lesson is that narrative can sustain a valuation only as long as it is not subjected to the scrutiny that fundamentals must withstand, and that a strategy which depends on the audience never examining the economics is not a strategy but a deferral of reckoning.

The Recurring Mechanisms and Their AI-Era Forms

The synthesis for the product leader is that these failures, despite their surface variety, recur through a small set of mechanisms that the AI transition is poised to reproduce. The refusal to cannibalize a profitable core, which destroyed Kodak, threatens every incumbent whose existing business an AI-native reconception would disrupt, since the organizational gravity that protected film now protects per-seat software and human-operated workflows against the agentic alternatives that would cannibalize them. The misreading of the basis of competition, which defeated Nokia and BlackBerry, threatens every company evaluating AI entrants against the old category definition, judging an AI product a poor version of the existing tool while it becomes a different and superior thing. The strategic indecision that dissipated Yahoo threatens every organization that responds to the AI transition with hedged half-measures rather than a committed diagnosis of what it must become. And the confusion of narrative with fundamentals that exposed WeWork threatens the many AI ventures whose valuations rest on stories that the eventual scrutiny of unit economics, particularly the inference costs examined earlier, may not support. The agenda for the product leader is to study these mechanisms not as historical curiosities but as live hazards, and to ask honestly which of them the organization is currently enacting, since the companies that failed did not believe they were making these mistakes either.

References

iDeals. (2023). WeWork IPO failure: Causes, collapse, and aftermath. https://www.idealsvdr.com/blog/deals/need-know-wework-ipo-postponement/

Killer Innovations. (2023). 5 innovation blind spots that killed Nokia and Kodak. https://killerinnovations.com/5-innovation-blind-spots-that-killed-nokia-and-kodak-s11-ep9/

Rumelt, R. P. (2011). Good strategy / bad strategy: The difference and why it matters. Crown Business.

The Product Operating Model and Its First Principles

The product operating model, as Cagan and colleagues (2024) develop it, is built on a small set of first principles that distinguish how leading product companies operate from how most organizations operate, and the central principle is that teams should be assigned problems to solve rather than features to build, with the responsibility and the authority to determine the best solution. The consequences of this principle propagate through the entire organization, since a model that empowers teams to solve problems requires that strategy be set centrally as a clear articulation of which problems matter and why, while solutions are developed locally by the teams closest to the customer and the technology, which is a specific division of labor between leadership and teams rather than a vague aspiration toward empowerment. The contrasting model, in which leadership specifies the features and teams implement them, fails to produce strategy in any meaningful sense, because it relegates teams to execution and concentrates all problem-solving in a leadership layer that lacks the proximity to customers and technology that good solutions require, with the result that the organization moves quickly in directions that are frequently wrong.

The strategic significance of the operating model for the product leader is that it determines whether the organization can learn, since a model that empowers teams to solve problems creates thousands of points at which the organization tests its assumptions against reality and adapts, whereas a model that dictates features from the center creates a single point of strategic reasoning whose errors propagate unchecked. The operating model is, in this sense, the organization’s learning architecture, and its design determines how quickly the organization discovers and corrects its mistakes.

Platform Teams and the Leverage of Shared Capability

Platform teams occupy a distinctive position in the operating model, since their purpose is not to deliver value to external customers directly but to provide internal capabilities that multiply the effectiveness of the teams that do, and the strategic logic of investing in platforms is that a capability built once and used by many teams produces leverage that no single team could achieve alone. The discipline of platform strategy is to treat the platform’s internal users, namely the product teams that depend on it, as genuine customers whose productivity the platform exists to serve, which means the platform must be designed for adoption and usability rather than imposed as a mandate, since a platform that internal teams route around because it is harder to use than building their own provides negative leverage. The recurring failure of platform strategy is to build platforms that serve the convenience of the platform team or the architectural preferences of leadership rather than the genuine needs of the teams meant to use them, and the corrective is to hold platform teams accountable for the adoption and the productivity gains of their internal customers rather than for the existence of the platform itself.

Centralized Versus Decentralized and the False Binary

The tension between centralized and decentralized product organizations is among the most persistent in the discipline, and it is frequently posed as a binary choice when the more accurate framing is a question of what to centralize and what to decentralize, since the two are not alternatives but complementary aspects of a coherent operating model. The pattern that the product operating model implies is the centralization of strategy and the decentralization of execution, in which leadership centrally sets the strategic context, namely the vision, the priorities, and the problems that matter, while teams decentrally develop the solutions, which combines the coherence that only central strategy can provide with the adaptiveness that only local problem-solving can achieve. The failure modes lie at both extremes: an organization that centralizes everything produces coherent strategy that cannot adapt and teams that cannot learn, while an organization that decentralizes everything produces adaptive teams whose efforts do not aggregate into a coherent direction, and the operating model’s task is to find the division that captures coherence and adaptiveness together rather than sacrificing one for the other.

The appropriate division shifts with the organization’s scale and context, since a small organization may need little formal central strategy because alignment is maintained through proximity, while a large organization requires explicit central strategy to prevent its many teams from diverging, and the product leader’s task is to design the operating model for the organization’s actual scale rather than importing the model of a company at a different stage.

Operating Models in the Age of AI

The AI transition tests operating models in ways the product leader should anticipate. The first test is to the pace of learning, since AI both accelerates the rate at which teams can build and test solutions and raises the uncertainty about which solutions will work, which increases the value of an operating model that empowers teams to learn quickly and penalizes a model that routes every decision through a slow central authority, widening the gap between the empowered and the feature-factory models. The second test concerns the platform layer specifically, since the infrastructure for building, deploying, evaluating, and governing AI is a substantial new platform responsibility that organizations must locate within their operating model, and the choice of whether to centralize this capability in a platform team or distribute it across product teams is among the more consequential operating-model decisions of the current moment. The third and most profound test is that agentic systems raise the prospect of an operating model in which some decisions and some execution are performed by autonomous systems rather than by human teams, which will require organizations to determine which decisions can be delegated to agents, how those agents are governed, and how human and agent work is coordinated, questions that existing operating models were not designed to address and that will define the next decade of organizational design.

The synthesis that closes both this module and the series is that the operating model is the deepest determinant of whether a product strategy can succeed, because how a company decides becomes what it builds. The agenda for the product leader is to build an operating model that empowers teams to solve problems rather than implement features, since this is the precondition for genuine strategy and organizational learning; to invest in platform teams that provide leverage while holding them accountable to their internal customers; to resolve the centralization question by centralizing strategy and decentralizing execution rather than treating the two as a binary; and to anticipate that the AI transition is widening the advantage of empowered learning models, creating a major new platform responsibility, and beginning to introduce autonomous systems into the model itself. A product leader who masters the operating model shapes the conditions under which every other strategic decision is made, which is the highest leverage available in the discipline, and the appropriate place for this series on the foundations of product strategy to conclude.

References

Cagan, M., Hickman, L., Jones, C., Idiodi, C., & Moore, J. (2024). Transformed: Moving to the product operating model. Wiley.

Skelton, M., & Pais, M. (2019). Team topologies: Organizing business and technology teams for fast flow. IT Revolution Press.

The Product Operating Model and Its First Principles

The product operating model, as Cagan and colleagues (2024) develop it, is built on a small set of first principles that distinguish how leading product companies operate from how most organizations operate, and the central principle is that teams should be assigned problems to solve rather than features to build, with the responsibility and the authority to determine the best solution. The consequences of this principle propagate through the entire organization, since a model that empowers teams to solve problems requires that strategy be set centrally as a clear articulation of which problems matter and why, while solutions are developed locally by the teams closest to the customer and the technology, which is a specific division of labor between leadership and teams rather than a vague aspiration toward empowerment. The contrasting model, in which leadership specifies the features and teams implement them, fails to produce strategy in any meaningful sense, because it relegates teams to execution and concentrates all problem-solving in a leadership layer that lacks the proximity to customers and technology that good solutions require, with the result that the organization moves quickly in directions that are frequently wrong.

The strategic significance of the operating model for the product leader is that it determines whether the organization can learn, since a model that empowers teams to solve problems creates thousands of points at which the organization tests its assumptions against reality and adapts, whereas a model that dictates features from the center creates a single point of strategic reasoning whose errors propagate unchecked. The operating model is, in this sense, the organization’s learning architecture, and its design determines how quickly the organization discovers and corrects its mistakes.

Platform Teams and the Leverage of Shared Capability

Platform teams occupy a distinctive position in the operating model, since their purpose is not to deliver value to external customers directly but to provide internal capabilities that multiply the effectiveness of the teams that do, and the strategic logic of investing in platforms is that a capability built once and used by many teams produces leverage that no single team could achieve alone. The discipline of platform strategy is to treat the platform’s internal users, namely the product teams that depend on it, as genuine customers whose productivity the platform exists to serve, which means the platform must be designed for adoption and usability rather than imposed as a mandate, since a platform that internal teams route around because it is harder to use than building their own provides negative leverage. The recurring failure of platform strategy is to build platforms that serve the convenience of the platform team or the architectural preferences of leadership rather than the genuine needs of the teams meant to use them, and the corrective is to hold platform teams accountable for the adoption and the productivity gains of their internal customers rather than for the existence of the platform itself.

Centralized Versus Decentralized and the False Binary

The tension between centralized and decentralized product organizations is among the most persistent in the discipline, and it is frequently posed as a binary choice when the more accurate framing is a question of what to centralize and what to decentralize, since the two are not alternatives but complementary aspects of a coherent operating model. The pattern that the product operating model implies is the centralization of strategy and the decentralization of execution, in which leadership centrally sets the strategic context, namely the vision, the priorities, and the problems that matter, while teams decentrally develop the solutions, which combines the coherence that only central strategy can provide with the adaptiveness that only local problem-solving can achieve. The failure modes lie at both extremes: an organization that centralizes everything produces coherent strategy that cannot adapt and teams that cannot learn, while an organization that decentralizes everything produces adaptive teams whose efforts do not aggregate into a coherent direction, and the operating model’s task is to find the division that captures coherence and adaptiveness together rather than sacrificing one for the other.

The appropriate division shifts with the organization’s scale and context, since a small organization may need little formal central strategy because alignment is maintained through proximity, while a large organization requires explicit central strategy to prevent its many teams from diverging, and the product leader’s task is to design the operating model for the organization’s actual scale rather than importing the model of a company at a different stage.

Operating Models in the Age of AI

The AI transition tests operating models in ways the product leader should anticipate. The first test is to the pace of learning, since AI both accelerates the rate at which teams can build and test solutions and raises the uncertainty about which solutions will work, which increases the value of an operating model that empowers teams to learn quickly and penalizes a model that routes every decision through a slow central authority, widening the gap between the empowered and the feature-factory models. The second test concerns the platform layer specifically, since the infrastructure for building, deploying, evaluating, and governing AI is a substantial new platform responsibility that organizations must locate within their operating model, and the choice of whether to centralize this capability in a platform team or distribute it across product teams is among the more consequential operating-model decisions of the current moment. The third and most profound test is that agentic systems raise the prospect of an operating model in which some decisions and some execution are performed by autonomous systems rather than by human teams, which will require organizations to determine which decisions can be delegated to agents, how those agents are governed, and how human and agent work is coordinated, questions that existing operating models were not designed to address and that will define the next decade of organizational design.

The synthesis that closes both this module and the series is that the operating model is the deepest determinant of whether a product strategy can succeed, because how a company decides becomes what it builds. The agenda for the product leader is to build an operating model that empowers teams to solve problems rather than implement features, since this is the precondition for genuine strategy and organizational learning; to invest in platform teams that provide leverage while holding them accountable to their internal customers; to resolve the centralization question by centralizing strategy and decentralizing execution rather than treating the two as a binary; and to anticipate that the AI transition is widening the advantage of empowered learning models, creating a major new platform responsibility, and beginning to introduce autonomous systems into the model itself. A product leader who masters the operating model shapes the conditions under which every other strategic decision is made, which is the highest leverage available in the discipline, and the appropriate place for this series on the foundations of product strategy to conclude.

References

Cagan, M., Hickman, L., Jones, C., Idiodi, C., & Moore, J. (2024). Transformed: Moving to the product operating model. Wiley.

Skelton, M., & Pais, M. (2019). Team topologies: Organizing business and technology teams for fast flow. IT Revolution Press.

Team Topology and the Management of Cognitive Load

The most rigorous contemporary treatment of team structure is the work of Matthew Skelton and Manuel Pais (2019), who argue that effective organizations are composed of a small number of fundamental team types with deliberately designed interactions, rather than an undifferentiated collection of feature teams. They distinguish stream-aligned teams that own a continuous flow of value to a specific customer segment or product area, platform teams that provide internal capabilities which accelerate the stream-aligned teams, enabling teams that temporarily build capability in other teams and then withdraw, and complicated-subsystem teams that own components requiring deep specialist expertise (Skelton & Pais, 2019). The strategic insight underlying this taxonomy is that the central constraint on a team’s effectiveness is its cognitive load, the total amount the team must hold in mind to do its work well, and that the purpose of the structure is to bound each team’s cognitive load to a sustainable level by allocating responsibilities so that no team is asked to understand more than it can.

The implication for the product leader is that team design is the design of cognitive load distribution, and that the common pattern of asking a single team to own an ever-expanding surface area is a structural error that degrades the team’s effectiveness regardless of the team’s talent. The corrective is to recognize when a team’s cognitive load has grown beyond sustainable bounds and to restructure, often by extracting a platform or complicated-subsystem team that absorbs the specialist complexity, which restores the stream-aligned team’s capacity to focus on customer value.

Ownership Models and the Conditions of Accountability

Ownership models, the assignment of responsibility for outcomes to teams, determine whether accountability is real or nominal, and the distinction turns on whether a team owns a coherent outcome it can actually influence or merely a fragment whose results depend on factors beyond its control. The most effective ownership model assigns a team a durable, meaningful slice of the product such that the team can reason about, influence, and be held accountable for a customer or business outcome, which is the structural precondition for the empowerment that high-functioning product organizations depend upon. The failure mode is fragmented ownership, in which responsibility for an outcome is distributed across so many teams that no single team can be accountable for it, with the consequence that the outcome is owned by everyone and therefore by no one, and the product leader’s task is to draw ownership boundaries around outcomes that teams can genuinely own rather than around components that happen to be technically separable.

The relationship between ownership and topology is intimate, since durable ownership of a customer outcome is precisely what the stream-aligned team is designed to provide, while the platform and complicated-subsystem teams own the capabilities and components that would otherwise fragment the stream-aligned team’s ownership, which is why topology and ownership must be designed together rather than separately.

Incentive Alignment and the Behaviors Structure Rewards

Incentive alignment, the configuration of how teams are measured and rewarded, is the force that determines what teams actually do regardless of what they are nominally asked to do, since teams optimize for what they are measured on with a reliability that overrides stated intentions. The strategic discipline of incentive alignment is to ensure that the outcomes teams are measured on are the outcomes the strategy actually requires, which is harder than it appears, because the metrics that are easy to measure and assign are frequently outputs such as features shipped rather than outcomes such as customer value delivered, and a team measured on output will produce output whether or not it produces value. The corrective, consistent with the empowered-team model, is to measure teams on the outcomes they own rather than on the output they produce, which aligns the team’s optimization with the strategy’s intent, though this requires the harder work of defining and measuring outcomes that a team can genuinely be held accountable for.

A particular incentive hazard is the misalignment between teams whose cooperation the strategy requires but whose individual incentives reward local optimization at the expense of the whole, which produces the familiar pathology of teams each succeeding on their own metrics while the product as a whole fails, and the product leader’s task is to detect and correct these misalignments by ensuring that the incentives of interdependent teams reward the shared outcome rather than only their separate contributions.

Building Strategic Teams in the Age of AI

The AI transition is altering team composition and structure in ways the product leader must anticipate. The first alteration is that AI tooling is changing the cognitive load and the leverage of individual teams, since capabilities that previously required dedicated specialist teams may increasingly be accessible to stream-aligned teams through AI assistance, which shifts where the topology’s boundaries should be drawn and may reduce the number of teams required to own a given surface. The second is the emergence of a new platform responsibility specific to AI, since the infrastructure for deploying, evaluating, and governing models is precisely the kind of specialist complexity that a platform team should own on behalf of stream-aligned teams, sparing them the cognitive load of becoming machine-learning specialists while giving them reliable access to AI capability. The third and most forward-looking is that as agentic systems take on portions of the work teams perform, the unit being organized may increasingly include autonomous agents alongside humans, which raises genuinely new questions about ownership and accountability that the existing frameworks were not designed to answer and that product leaders will need to work out in practice.

The synthesis for the product leader is that structure is strategy in disguise, since the team structure determines what the organization can build and how its products will be shaped, by Conway’s logic. The agenda is to design team topology deliberately as the management of cognitive load, extracting platform and specialist teams to keep stream-aligned teams focused on customer value; to assign ownership of coherent outcomes that teams can genuinely influence rather than fragments that diffuse accountability; to align incentives so that teams are measured on the outcomes the strategy requires rather than the output that is easy to count; and to anticipate that the AI transition is reshaping cognitive load, creating a new platform responsibility around models, and beginning to introduce agents as members of the teams being organized. A product leader who shapes structure with the same intentionality applied to roadmaps will find that many problems attributed to execution were structural all along, and that the most durable strategic interventions are frequently changes to who owns what rather than changes to what gets built.

References

Cagan, M., & Jones, C. (2020). Empowered: Ordinary people, extraordinary products. Wiley.

Skelton, M., & Pais, M. (2019). Team topologies: Organizing business and technology teams for fast flow. IT Revolution Press.

The Narrative Memo and the Discipline of Prose

The narrative memo, a document of structured prose rather than bulleted slides, was elevated to an organizational practice most famously at Amazon, where Jeff Bezos instituted a requirement that proposals be presented as written narratives rather than presentations, having concluded after considerable experimentation that the narrative form forced a quality of thought the slide deck permitted teams to evade (Bryar & Carr, 2021). The mechanism is specific and worth stating precisely: a slide deck allows ideas to be presented as fragments whose logical relationships and relative importance remain unstated, which permits the author to gloss over weak connections and the audience to fill gaps with charitable assumptions, whereas prose requires the author to state how each idea relates to the next, which exposes the places where the reasoning does not actually connect. The discipline of writing a coherent six-page narrative is harder than producing a twenty-slide deck precisely because the narrative cannot hide its gaps, and the difficulty is the point, since the document that is hard to write is the document whose weaknesses were found and fixed before the strategy was executed rather than after.

The organizational practice that surrounds the narrative memo, in which the document is read in silence at the start of a meeting before discussion begins, reinforces its value by ensuring that the audience engages with the full argument rather than with a presenter’s selective narration of it, which subjects the strategy to genuine scrutiny rather than to the persuasive performance that slides invite.

The Strategy Document and the Avoidance of the Bad-Strategy Trap

The strategy document, the durable articulation of where a product is going and why, earns its value when it avoids what Richard Rumelt (2011) terms bad strategy, which he characterizes as the substitution of ambitious goals, vision statements, and aspirations for the actual content of strategy, namely a diagnosis of the situation, a guiding policy for addressing it, and a coherent set of actions that follow. Rumelt’s contribution is to insist that a real strategy contains a clear-eyed diagnosis of the challenge the organization actually faces, since a strategy that does not name the central obstacle cannot coherently address it, and the most common failure of strategy documents is that they enumerate goals and initiatives without ever diagnosing the problem those initiatives are meant to solve. A strategy document built on Rumelt’s structure forces the product leader to state what is genuinely hard about the situation before proposing what to do about it, which prevents the document from becoming the list of aspirations and disconnected initiatives that masquerades as strategy in most organizations.

The Press-Release-and-FAQ Format and Working Backwards

The press-release-and-FAQ format, developed at Amazon as the instrument of its working-backwards process, addresses a different failure, which is the tendency to design a product around what is feasible to build rather than around what is valuable to the customer (Bryar & Carr, 2021). The format requires the team to write, before building anything, the press release that would announce the finished product as though it were launching, articulating in customer-facing language the value the product delivers and why the customer should care, followed by the frequently asked questions that surface and answer the hard problems the product must solve. The discipline this imposes is to begin from the customer experience and reason backward to what must be built, rather than beginning from what can be built and reasoning forward to a customer who might want it, which inverts the failure mode that produces technically impressive products no customer values. The format also functions as a forcing device for honesty, since a press release that does not describe a compelling customer benefit reveals, before any resources are committed, that the idea is not yet worth building.

The One-Page Narrative and the Compression of Conviction

The one-page narrative, the discipline of compressing a strategy into a single page, is the most demanding format because compression is where understanding is tested most severely, as a strategy that cannot be expressed clearly in one page is frequently a strategy that is not yet clearly understood. The value of the one-pager is not that it is convenient for busy readers, though it is, but that the act of compression forces the author to identify what is genuinely essential and to discard what is merely included, which clarifies the strategy for the author before it clarifies it for any reader. The one-page narrative also serves the propagation function described in the prior module, since a strategy compressed to its essence is portable in a way that a long document is not, traveling through the organization in the form that people can hold in memory and repeat accurately.

Strategy Communication in the Age of AI

The AI transition acts on strategy communication in two directions the product leader should weigh. The first is that AI can now draft, structure, and critique strategy documents, which is genuinely useful for overcoming the blank page and for stress-testing an argument against anticipated objections, but which carries a specific hazard, since the value of these documents was never the prose but the thinking the prose compelled, and a strategy document generated rather than reasoned through delivers the artifact while skipping the cognitive work that was its entire purpose. The product leader must therefore use AI to sharpen and pressure-test thinking that the leader has actually done, rather than to substitute for the thinking, since a fluent document produced without the underlying reasoning is more dangerous than an awkward one produced with it, as it carries the appearance of rigor without its substance. The second is that strategy documents in the AI era must increasingly address questions that did not previously arise, namely how the strategy accounts for the commoditization of capability, where its defensibility resides, and how it positions against an uncertain trajectory of model advancement, which means the content these frameworks must carry has expanded even as the tools for producing them have improved.

The synthesis for the product leader is that strategy communication frameworks are thinking tools first and communication tools second, and that their discipline is the source of their value. The agenda is to write strategy as narrative prose that exposes the gaps slides conceal, to structure strategy documents around diagnosis rather than aspiration, to use the working-backwards format to reason from customer value rather than from feasibility, to compress strategy to a single page as a test of genuine understanding, and to deploy AI as an instrument that sharpens reasoning the leader has done rather than as a substitute for the reasoning itself. The document is the thinking, and the product leader who treats the writing as the work, rather than as the record of work done elsewhere, will produce both better strategies and better organizational alignment around them.

References

Bryar, C., & Carr, B. (2021). Working backwards: Insights, stories, and secrets from inside Amazon. St. Martin’s Press.

Rumelt, R. P. (2011). Good strategy / bad strategy: The difference and why it matters. Crown Business.

Executive Communication and the Economy of Attention

Executive communication is governed by a constraint that junior product managers consistently underestimate, which is that senior leaders allocate attention with extreme scarcity and decide whether to engage with an idea in the first moments of exposure to it. The implication is that the product leader’s communication must be structured for a reader who is impatient, skeptical, and reasoning at a higher level of abstraction than the product details that occupy the team, which means leading with the conclusion and the stakes rather than building toward them, and framing the idea in terms of the outcomes the executive is accountable for rather than the features the team is proud of. The mechanism by which this earns influence is that it respects the executive’s economy of attention, demonstrating that the product leader has already done the work of distilling what matters, which builds the credibility that causes the executive to extend trust to the leader’s judgment on the details they did not have time to examine.

The deeper discipline of executive communication is to translate between levels of abstraction fluently, since the product leader sits between an executive layer reasoning in terms of markets, financials, and strategic bets and a team layer reasoning in terms of features, experiments, and technical constraints, and the leader’s value is partly the ability to carry meaning faithfully across that gap, expressing team-level realities in executive-level terms and executive-level strategy in team-level direction. A product leader who can only speak in one register is confined to influence within that register, while one who translates between them becomes the connective tissue through which strategy actually propagates.

Narrative Building: The Construction of Shared Conviction

Narrative building is the work of assembling facts, analysis, and judgment into a coherent account of why a particular strategic direction is correct, and its power derives from a property of human cognition that pure analysis lacks, which is that people remember, repeat, and act upon stories far more reliably than they do upon disconnected facts or recommendations. A strong narrative does not merely present a conclusion but constructs the reasoning that makes the conclusion feel inevitable, beginning from a situation the audience recognizes, identifying the tension or opportunity within it, and arriving at the proposed direction as the natural resolution of that tension. The strategic value of narrative is that it is portable, since a narrative that genuinely persuades does not require the product leader’s continued presence to spread, but is carried by the people who adopt it into rooms the leader will never enter, which is how a product leader’s influence scales beyond the meetings they personally attend.

The construction of a durable narrative depends on intellectual honesty as much as on rhetorical skill, since a narrative that suppresses inconvenient facts or overstates certainty will be discovered and will damage the leader’s credibility, whereas a narrative that acknowledges the genuine uncertainties and the strongest objections, and addresses them, earns the trust that makes it persuasive to sophisticated audiences. The product leader’s task is therefore to build narratives that are both compelling and true, which is harder than building narratives that are merely compelling, and which is the only kind that survives contact with the skeptical executives whose conviction the leader needs.

Strategic Storytelling and the Anchoring of Abstraction

Strategic storytelling, the use of concrete, specific narratives to make abstract strategy tangible, is the technique that converts a strategy from a proposition the audience evaluates into an experience the audience inhabits. The reason this matters is that strategy is inherently abstract, dealing in markets, segments, and bets that the audience cannot directly perceive, and abstraction is difficult to believe and easy to dispute, whereas a concrete story about a specific customer, a specific moment of value, or a specific future state gives the audience something to hold and to feel, which engages conviction in a way that abstraction does not. The most effective product leaders anchor their strategies in vivid, specific scenes, describing the experience of a named customer using the future product or the precise moment at which the strategy delivers value, because these scenes make the strategy real to the audience before the analysis asks them to believe in it.

In the context of AI, strategic storytelling acquires unusual importance because the AI transition is genuinely difficult to reason about in the abstract, since its trajectory is uncertain and its implications are unfamiliar, which means that organizations are particularly prone to either dismissing it or pursuing it incoherently. The product leader who can tell a concrete, credible story about what the organization’s product becomes in an AI-shaped future, anchored in a specific customer experience rather than in generalities about transformation, provides the organization with something it badly needs, which is a tangible picture of a future it can otherwise only abstractly fear or vaguely anticipate, and this storytelling is among the most valuable forms of leadership available in the current moment.

The synthesis for the product leader is that strategic influence is a communication discipline as much as an analytical one, and that the leader who invests only in being right while neglecting the work of persuasion will watch better-communicated and lesser ideas prevail. The agenda is to structure executive communication for the economy of attention by leading with conclusions and outcomes, to build narratives that construct shared conviction through reasoning that is both compelling and honest, to anchor abstract strategy in concrete storytelling that the audience can inhabit, and to recognize that the AI transition has raised the premium on the leader who can make an uncertain future tangible. Influence without authority is the defining condition of product leadership, and the product leaders who master the persuasive work of moving an organization, rather than merely the analytical work of knowing what it should do, are the ones whose strategies are actually adopted.

References

Cagan, M., & Jones, C. (2020). Empowered: Ordinary people, extraordinary products. Wiley.

Duarte, N. (2010). Resonate: Present visual stories that transform audiences. Wiley.

Rumelt, R. P. (2011). Good strategy / bad strategy: The difference and why it matters. Crown Business.

Why Product, Sales, and Marketing Pull in Different Directions

The structural source of go-to-market misalignment is that product, sales, and marketing optimize for different objectives on different time horizons and are frequently compensated accordingly, which produces predictable and recurring tension. Product organizations tend to favor motions in which the product itself drives acquisition and expansion, since these motions reward exactly the product quality the team is building, whereas sales organizations frequently prefer larger contracts closed through human relationships, since these motions reward the activity sales is compensated for, and the two preferences point toward different products, different pricing, and different definitions of success (Sapphire Ventures, 2023). The most consequential decision that must be aligned is whether the dominant motion is product-led, in which users discover, try, and adopt the product with minimal human involvement, or sales-led, in which a sales organization guides prospects through a considered purchase, because this choice propagates into nearly every other decision, from how the product onboards users to how it is packaged and priced.

The contemporary reality is that the question is rarely answered purely, since the prevailing pattern across successful software companies has become a hybrid in which a product-led motion acquires and activates users at the low end while a sales-led motion captures and expands enterprise accounts at the high end, which means the alignment task is not to choose one motion but to ensure the two operate as a coordinated system in which the product-led motion feeds qualified demand into the sales-led one rather than competing with it.

Positioning and the Collaboration with Product Marketing

Positioning, the articulation of what the product is, who it is for, and why it is the best available option for them, is the connective tissue that holds the go-to-market system together, and April Dunford (2019) makes the essential argument that positioning cannot live in the marketing department alone but requires the product, sales, and leadership functions in the room, because positioning that the product cannot deliver or that sales will not use is positioning that exists only on a slide. The collaboration between product management and product marketing is where positioning is forged, and its productive form is a genuine partnership in which product management supplies the deep understanding of what the product does and for whom, while product marketing supplies the discipline of translating that understanding into a position the market will recognize and the sales organization can carry. The recurring dysfunction is to treat product marketing as a downstream messaging function activated near launch, which wastes the function’s strategic value, since the most useful positioning work happens early enough to influence what the product becomes, not merely how it is described after it is finished.

Adoption Strategy as a Product Responsibility

Adoption strategy, the deliberate design of how a user moves from first contact to habitual, value-realizing use, is the area where product and go-to-market most directly converge, and it is increasingly a product responsibility rather than a marketing one. The reason is that in any product-led motion the product itself must perform the work of activation that a salesperson would otherwise perform, which means the onboarding experience, the time to first value, and the mechanisms that convert initial use into habit are product features that determine go-to-market success as directly as any marketing campaign. The strategic discipline is to identify the specific early behavior that predicts long-term retention, often termed the activation moment, and to design the product so that new users reach it as quickly and reliably as possible, since a product that reliably delivers its core value within the first session converts and retains at rates that no amount of downstream marketing can manufacture for a product that does not.

Go-To-Market in the Age of AI

The AI transition reshapes go-to-market alignment in three ways the product leader should anticipate. The first is that AI lowers the cost of the product-led motion by enabling products to onboard, guide, and support users through intelligent assistance that previously required human effort, which extends the reach of product-led motions into more complex products that formerly demanded sales involvement. The second is that AI compresses the time to value when designed well, since a product that can understand a user’s intent in natural language and act on it removes much of the configuration burden that delayed activation in conventional software, which strengthens adoption strategy directly. The third and most structurally novel is that the entity adopting the product may increasingly be an agent acting on a user’s behalf rather than a human navigating an interface, which means the product’s discoverability and adoption must extend to the protocols and interfaces through which autonomous systems find and invoke capabilities, a go-to-market surface that did not previously exist and that the product leader must now design for alongside the human one.

The synthesis for the product leader is that go-to-market alignment is a product responsibility because the product encodes the assumptions on which the entire motion rests. The agenda is to ensure that the product, its positioning, and its selling motion form one coherent system rather than three conflicting ones; to resolve the product-led versus sales-led question deliberately, typically as a coordinated hybrid rather than a pure choice; to partner with product marketing early enough that positioning shapes the product rather than merely describing it; to own adoption strategy as the product work of reliably delivering first value; and to anticipate that the AI transition both strengthens product-led adoption and introduces agents as a new class of adopter the product must reach. A product and the way it is sold that agree with each other compound their effectiveness, and a product and a go-to-market motion that disagree will, regardless of the quality of either in isolation, undermine each other in the market.

References

Dunford, A. (2019). Obviously awesome: How to nail product positioning so customers get it, buy it, love it. Ambient Press.

Moore, G. A. (2014). Crossing the chasm: Marketing and selling disruptive products to mainstream customers (3rd ed.). HarperBusiness.

Sapphire Ventures. (2023). Navigating product-led growth vs. sales-led growth models. https://sapphireventures.com/blog/navigating-product-led-growth-vs-sales-led-growth-models/

Customer Acquisition Cost and Lifetime Value: The Defining Ratio

The foundational relationship in unit economics is the ratio between the lifetime value of a customer and the cost of acquiring that customer, a relationship David Skok (2010) crystallized into the widely adopted heuristic that a healthy software business should generate roughly three times the lifetime value relative to acquisition cost, with the acquisition cost recovered comfortably within the first year. The logic of the ratio is straightforward: customer acquisition cost is the total sales and marketing expense divided by the customers it produced, and lifetime value is the gross margin a customer generates over the duration of the relationship, so the ratio expresses whether the firm earns substantially more from a customer than it spent to acquire one, which it must, since the surplus funds everything else the business does. A ratio near one means the firm is buying revenue at cost and cannot sustain itself, while a ratio far above three may indicate not health but under-investment in growth, which is why the heuristic targets a balance rather than a maximum.

The strategic point for the product leader is that both terms of the ratio are products of product decisions as much as of go-to-market spend, since a product that activates users quickly and demonstrates value reduces the acquisition cost by converting trials efficiently, and a product that retains and expands customers raises lifetime value directly. Treating the ratio as a marketing metric rather than a product one is the error that severs product strategy from the economics it determines.

Margins and the AI Disruption of Software’s Best Feature

The defining economic advantage of software has historically been its margin, since the marginal cost of serving an additional customer was close to zero, which allowed mature software businesses to operate at gross margins in the range of seventy-five to eighty percent and to convert revenue growth into profit at a rate few other industries could match. The AI transition disturbs this advantage at its foundation, because AI capability carries a real and recurring marginal cost in the form of inference, and analysis of AI-native companies suggests their gross margins run materially below the software norm, with estimates placing many LLM-native businesses near the low fifties rather than the high seventies (Foundry CRO, 2026). The strategic consequence is severe and underappreciated: a lower gross margin reduces the lifetime value term of the central ratio, which means an AI-native company must achieve a higher revenue-to-acquisition-cost ratio than a conventional software business to be equally efficient, since more of its revenue is consumed by the cost of serving the customer rather than retained as margin. A product leader who imports the unit-economics assumptions of conventional software into an AI product will systematically overestimate the health of the business, and the corrective is to model margin explicitly, treating inference cost as a product variable to be managed through model selection, caching, routing, and the deliberate design of which workloads invoke expensive capability.

Retention Economics: The Highest-Leverage Variable

Beneath both acquisition cost and lifetime value sits retention, which is the variable that most powerfully determines whether the economics compound or decay, because retention enters the lifetime value calculation multiplicatively rather than additively. A reduction in churn extends the duration over which every customer generates margin, and analysis indicates that even a modest reduction in churn produces a disproportionately larger improvement in lifetime value, since the effect accrues across the entire customer base and across the entire extended lifetime (Skok, 2010). The further refinement is net revenue retention, which captures not only the customers retained but the revenue expanded from them, and businesses that sustain net revenue retention above one hundred percent grow from their existing base alone, which is the most capital-efficient growth available and the reason markets reward it so richly. The product implication is that retention and expansion are not customer-success concerns to be addressed after the product is built but design objectives that the product must serve directly, since a product that deepens its value as the customer’s usage grows produces the expansion that drives net revenue retention, while a product whose value plateaus produces the churn that erodes lifetime value.

The AI Transition’s Double Effect on Unit Economics

The AI transition acts on unit economics in two opposing directions that the product leader must hold together. On the cost side, as already noted, inference compresses margin and therefore reduces lifetime value, which raises the efficiency the rest of the business must achieve to compensate. On the value side, however, AI can improve the other terms of the equation, since AI-native products that genuinely automate valuable work can command higher prices and deeper retention than the tools they replace, and AI can reduce acquisition cost by improving activation and by enabling product-led motions that convert users without expensive sales effort. The net effect on any specific product is therefore not predetermined but depends on whether the value AI adds to price and retention exceeds the margin it removes through inference cost, which is precisely the calculation the product leader must perform rather than assume. The products that will prove durable are those whose AI capability adds more to willingness to pay and to retention than it subtracts from margin, and identifying whether a product is on the right side of this calculation is among the most important analyses a product leader can conduct.

The synthesis is that unit economics is not finance’s concern imposed upon product but product strategy expressed in arithmetic. The agenda for the product leader is to own both terms of the acquisition-cost-to-lifetime-value ratio as product outcomes, to model margin explicitly rather than assuming software’s historical generosity, to treat retention and expansion as primary design objectives given their multiplicative effect on lifetime value, and to perform deliberately the calculation of whether AI’s contribution to price and retention exceeds its cost to margin. A strategy that survives this arithmetic can pursue any vision it chooses, and a strategy that ignores it will fail regardless of how compelling the vision appears.

References

Foundry CRO. (2026). LTV:CAC ratio benchmarks 2026. https://foundrycro.com/blog/ltv-cac-ratio-benchmarks-2026/

Gurley, B. (2011). All revenue is not created equal: The keys to the 10X revenue club. Above the Crowd.

https://abovethecrowd.com

Skok, D. (2010). SaaS metrics 2.0: A guide to measuring and improving what matters. For Entrepreneurs. https://www.forentrepreneurs.com/saas-metrics-2/

Value-Based Pricing and the Willingness-to-Pay Conversation

Value-based pricing sets price according to the value the product delivers to the customer rather than the cost of producing it or the prices competitors charge, and its strategic superiority follows from a simple observation, which is that cost-plus pricing leaves value on the table whenever the product is worth more than it costs, and competitor-matched pricing surrenders the firm’s own value proposition to the market’s lowest common denominator. Ramanujam and Tacke (2016) argue that the foundational discipline is to have the willingness-to-pay conversation with customers early, before the product is built, so that the team learns which capabilities customers value enough to pay for and designs accordingly, rather than building broadly and discovering later that customers will not pay for much of what was built. The mechanism by which this conversation creates value is that it converts pricing from a guess into a designed outcome, allowing the team to concentrate investment on the features that command willingness to pay and to avoid the costly error of over-building features that impress internally but command no premium.

In the context of AI, value-based pricing acquires a new salience because AI features often deliver value that is dramatically disproportionate to their cost, which means cost-plus pricing of AI capability systematically under-captures, while at the same time the underlying inference cost is real and variable, which means pricing that ignores cost entirely risks negative margins on heavy users. The strategic response is to price AI capability on the value of the outcome it produces while metering the consumption that drives cost, holding both the value the customer receives and the cost the firm incurs in the same pricing design.

Packaging: The Architecture of Choice

Packaging, the configuration of features into tiers and bundles, is where much of pricing strategy is actually executed, and it is frequently underestimated as a mere presentation layer rather than recognized as a structural determinant of revenue. Ramanujam and Tacke (2016) frame good packaging as the design of a small number of offers that segment customers by willingness to pay, so that each customer self-selects into the package that matches the value they derive, which allows the firm to acquire price-sensitive customers at the low end while capturing the full willingness to pay of high-value customers at the high end. The discipline of packaging is to align the dimension along which packages differ, the value metric, with the dimension along which customers derive value, so that customers who get more value naturally land in higher tiers as they grow, which is the mechanism that connects packaging to expansion revenue.

The recurring packaging error is to offer too many options or to differentiate packages on dimensions customers do not care about, which produces confusion rather than clean self-selection, and the corrective is the deliberate simplicity that Ramanujam advocates, in which a few well-differentiated packages map cleanly to identifiable segments.

Expansion Revenue and the Primacy of Net Revenue Retention

The most consequential shift in how mature software businesses think about monetization is the elevation of expansion revenue, the additional revenue earned from existing customers over time, to a position of primacy over new acquisition. The metric that captures this is net revenue retention, which measures the revenue retained and expanded from a cohort of existing customers, and its strategic importance is that a business whose existing customers reliably spend more over time grows even without acquiring new ones, which compounds in a way that acquisition-dependent growth cannot. Industry analysis indicates that companies sustaining net revenue retention above one hundred percent grow markedly faster than those below it, and that the value markets assign rises sharply as retention climbs into the higher bands, which establishes expansion as among the highest-leverage objectives a product organization can pursue (OpenView, 2023). The product implication is that the product must be designed so that value, and therefore spending, grows with the customer’s success, which is precisely what a well-chosen usage metric or a well-structured tier ladder accomplishes, connecting packaging design directly to the retention economics examined in the next module.

Strategic Pricing Mistakes That Recur

Several pricing mistakes recur with enough regularity to warrant naming. The first is pricing on cost rather than value, which under-captures whenever the product is worth more than it costs and is especially destructive for AI products whose value far exceeds their inference cost. The second is leaving pricing until the product is built, which forecloses the willingness-to-pay conversation that should have shaped what was built and frequently results in a product replete with features customers will not pay for. The third is changing price too rarely, treating the initial price as permanent when willingness to pay and competitive context evolve, with the consequence that the firm under-prices for years out of inertia. The fourth, particularly acute in the AI transition, is pricing AI features as free additions to an existing subscription in order to drive adoption, which can convert a profitable product into an unprofitable one as inference costs accumulate against revenue that did not rise to meet them.

The synthesis for the product leader is that pricing is product strategy and must be owned as such rather than delegated downstream. The agenda is to have the willingness-to-pay conversation before building, designing the product around the value customers will pay for; to architect packaging as a deliberate segmentation that allows self-selection and enables expansion; to manage net revenue retention as a primary objective by ensuring spending grows with customer success; and to avoid the recurring mistakes, with particular vigilance toward the AI-specific trap of giving away capability whose marginal cost is real. In an era where the value of intelligence is high and its cost is no longer negligible, the firms that treat pricing as a core product discipline will capture the value they create, and those that treat it as an afterthought will create value for customers and competitors to enjoy.

References

OpenView. (2023). 2023 SaaS benchmarks report: Pricing, packaging, and net revenue retention.

https://openviewpartners.com

Ramanujam, M., & Tacke, G. (2016). Monetizing innovation: How smart companies design the product around the price. Wiley.

Subscription Software and the Per-Seat Assumption

The subscription software model, in which a customer pays a recurring fee for continued access, became the default for business software because it aligns the vendor’s incentive with sustained value delivery and produces the predictable, compounding revenue that markets reward. Its dominant packaging form has been the per-seat license, which charges by the number of users, and the strategic elegance of per-seat pricing is that it scales revenue with the customer’s adoption while remaining simple to forecast and to sell. The limitation this model conceals is that per-seat pricing assumes value scales with the number of human users, an assumption the AI transition is quietly invalidating, since when the work is performed by an agent rather than a seated human, the number of seats decouples from the value delivered, and a vendor whose revenue is tied to seats faces the prospect of delivering more value while billing for fewer users.

Usage-Based Pricing and Its AI-Driven Resurgence

Usage-based pricing, in which the customer pays in proportion to consumption, is the model that has gained the most ground recently, and its resurgence is inseparable from the economics of AI. Industry data indicates that usage-based pricing reached mainstream adoption among software companies, and that hybrid arrangements combining a base subscription with a usage component have become the prevailing pattern rather than the exception (OpenView, 2023). The reason the AI transition forces this shift is mechanistic: AI features carry a real marginal cost in the form of inference, which means that unlike conventional software whose marginal cost is near zero, an AI product that prices on a flat subscription exposes itself to customers whose heavy usage destroys the unit economics. Usage-based pricing realigns the meter with the cost, charging more when the customer consumes more of the expensive resource, which is why the model has moved from a niche choice to a near-necessity for AI-native products. The strategic implication for the PM is that the pricing meter is now a product decision, since the product must measure, expose, and shape the consumption it bills for, which is a design responsibility that did not exist under flat subscription.

Marketplaces, Freemium, and Advertising

The marketplace model, in which the business captures a fee on transactions it facilitates between independent buyers and sellers, monetizes the network effect examined in an earlier module and carries the cold-start difficulty that defines all network businesses, since the model generates revenue only once both sides are present in density. Its strategic appeal is that the marketplace need not produce the underlying value itself, capturing instead a share of value created by participants, which can produce extraordinary margins at scale, though it demands the patience to solve the chicken-and-egg liquidity problem first.

Freemium, in which a capable free tier acquires users who convert to paid tiers for advanced capability, is a customer-acquisition model as much as a monetization model, and its discipline lies in the placement of the boundary between free and paid, which must be generous enough to demonstrate value yet constrained enough to motivate upgrade. The AI transition complicates freemium specifically because the free tier now carries inference cost, which means the generous free tier that acquired users at near-zero marginal cost in conventional software can become a significant and unbounded expense in an AI product, forcing a more deliberate design of what the free tier may consume.

The advertising-supported model, in which attention is monetized by selling access to it, decouples the user from the payer, which has the strategic consequence that the product is optimized for engagement and attention rather than for the user’s own stated goals, a tension that has defined the model’s history. In the context of AI, advertising support faces a structural question, since AI assistants that complete a task directly may bypass the surfaces on which advertising depends, which threatens the attention that the model monetizes.

Platform Economics as the Meta-Model

Platform economics is less a separate model than a meta-model that can sit atop the others, describing the situation in which a product captures value by orchestrating an ecosystem of third parties rather than by producing all value itself. The strategic distinction of platform economics is that the platform’s marginal cost of supporting an additional participant is low while the value created by additional participants is high, which produces the increasing returns to scale that make platforms so formidable when they succeed and so difficult to bootstrap before they do. The AI transition raises the prospect of a new platform layer in which the participants are not only human developers but autonomous agents, and the platform that becomes the venue where agents discover and transact for capabilities would occupy a position analogous to the dominant platforms of the prior era.

Choosing and Combining Models in the Age of AI

The synthesis for the product leader is that business models are not given but chosen, and that the choice is increasingly a hybrid construction rather than a selection of one model from a menu. The prevailing pattern across the most successful software companies is the combination of a subscription floor that provides predictability, a usage component that aligns revenue with the cost and value of consumption, and frequently a freemium or platform layer that drives acquisition or ecosystem growth, assembled deliberately to match the product’s economics and its customers’ willingness to pay. The AI transition reorders this calculus by introducing genuine marginal cost, by decoupling value from human seats, and by threatening the attention and discovery surfaces that several models relied upon, which means that a PM building an AI product who imports the business model of the conventional product it resembles is likely importing an economics that no longer holds. The discipline this module sets is to treat the business model as a first-class product decision, to understand the behaviors and metrics each model rewards, and to reason explicitly about how the marginal cost of intelligence and the migration of work to agents change which model the product should adopt and how its meter should be designed.

References

Cusumano, M. A., Gawer, A., & Yoffie, D. B. (2019). The business of platforms: Strategy in the age of digital competition, innovation, and power. Harper Business.

OpenView. (2023). 2023 SaaS benchmarks report: Usage-based pricing and monetization trends.

https://openviewpartners.com

Parker, G. G., Van Alstyne, M. W., & Choudary, S. P. (2016). Platform revolution: How networked markets are transforming the economy and how to make them work for you. W. W. Norton.

AI-Native Versus AI-Enabled: A Distinction of Architecture, Not Marketing

The first distinction a product leader must reason about precisely, because the market uses it loosely, is between AI-enabled and AI-native products. An AI-enabled product is one to which AI has been added as a feature while the product’s fundamental architecture and value proposition would survive the removal of that feature, whereas an AI-native product is one whose value proposition collapses entirely without AI because the intelligence is constitutive of what the product is rather than additive to it (Andreessen Horowitz, 2024). The test is counterfactual and clarifying: if the product would function essentially as before with the AI removed, it is AI-enabled, and if its core reason for existing disappears, it is AI-native. The strategic significance of the distinction is that AI-enabled features, being additive, are the most easily replicated and therefore the least defensible, since a competitor can add the same feature to its own product, while AI-native products, being architecturally reorganized around intelligence, can pursue value propositions and workflows that incumbents bolting AI onto legacy architectures cannot match.

The deeper implication, which connects to the blue ocean and disruption logic of earlier modules, is that the most consequential AI opportunities are typically AI-native reconceptions of a category rather than AI-enabled enhancements of an existing product, because the native architecture can eliminate the assumptions the legacy product was built around. A product leader who frames the AI question as how to add AI features to the current product is, by the structure of the question, confined to the less defensible position, whereas one who asks what the product would be if rebuilt around intelligence from the foundation is positioned to find the durable opportunity.

Agentic Systems: From Assistance to Execution

The most significant capability shift within the AI transition is the move from systems that assist a human performing a task to systems that execute the task autonomously, which is the domain of agentic products. An agentic system is one that pursues a goal through a sequence of actions, invoking tools, making decisions, and adapting to results with limited human intervention, and the strategic consequence of this shift is that the unit of value moves from augmenting human productivity to performing the work itself. This relocates the competitive battleground in two directions at once, as the analysis of the current environment suggests: defensibility shifts downward toward the data models, permissions, workflow logic, and compliance infrastructure that autonomous execution requires, and upward toward the networks, proprietary data generation, and real-world execution that agents enable (Andreessen Horowitz, 2024).

For the product leader, the rise of agentic systems poses both the largest opportunity and the largest substitution threat reasoned about in Module 10, since a sufficiently capable general-purpose agent is the substitute that can absorb the job a specialized product performs. The strategic response is to determine whether the product is positioned to be the agent that executes the valuable workflow, in which case the opportunity is to own end-to-end execution, or whether it is positioned to be a capability the agent invokes, in which case the opportunity is to become an indispensable tool in the agent’s repertoire, accessible through the interfaces that autonomous systems use. The error is to assume the product can remain a human-operated interface in a domain where execution is migrating to agents, since that is the position most exposed to substitution.

Commoditization Risk: The Defining Condition of the Environment

The condition that organizes all AI strategy is that foundation model capability is commoditizing, which means that the raw intelligence a product can access is becoming abundant, broadly available, and decreasingly differentiated across providers and over time. The strategic consequence, stated bluntly by the analysis emerging from the venture community, is that using a capable model or adding an agent is easy to replicate and therefore is not a moat, however impressive the capability may be (Andreessen Horowitz, 2024). A product whose differentiation rests solely on access to model capability is building on a commodity, and as the commodity becomes more capable and more widely available, that differentiation erodes toward zero, which is the fate awaiting the large population of products that are thin wrappers over a rented model. The product leader must internalize that the model is the engine but not the moat, and that strategy in this environment is precisely the work of building advantage that does not reside in the commoditizing layer.

This condition also reframes the supplier-power analysis of Module 10 into a defining strategic risk, since dependence on a commoditizing capability supplied by concentrated providers means that the value a product adds must be located somewhere the provider does not capture, lest the provider, by improving its base offering, absorb the product’s reason for existing. The history of platform shifts is replete with applications rendered obsolete when the platform incorporated their function, and the AI equivalent is the product whose entire value is a capability the next model release will provide natively.

Where Defensibility Actually Resides

The constructive core of AI strategy is the identification of the layers where advantage survives commoditization, and the convergent finding across the most credible analyses is that defensibility lives not in the model but in the assets and relationships that surround it. The first such asset is the proprietary data feedback loop, which becomes more valuable in an AI world rather than less, because when the model is the commodity the scarce input becomes the proprietary data that specializes it, and a product whose usage generates data no competitor can observe builds a model advantage that widens with scale (V7 Labs, 2024). This is the data network effect of Module 14 expressed as the central AI moat, and it is the single most reliable source of durable advantage available to an AI product.

The second is workflow ownership and embedding, the depth to which a product is woven into the systems of record, the daily practices, and the surrounding tools of its customers, since a product that owns an end-to-end workflow and serves as the system of record accrues switching costs that a competitor renting the same model cannot overcome (Andreessen Horowitz, 2024). The third is the network effect among the product’s users, which an AI agent cannot replace, because the value that arises from the presence of other people, as in collaboration and communication products, is not a capability a model can synthesize away (Bain & Company, 2025). The fourth is trust and the surrounding evaluation, governance, and reliability infrastructure, which becomes a durable differentiator precisely because AI outputs are probabilistic and consequential, so that the product which has earned trust and built the mechanisms that sustain it holds an advantage that is slow to accumulate and therefore slow for competitors to replicate.

A sober note belongs here, since the enterprise reality tempers the enthusiasm: a substantial fraction of agentic projects are forecast to be abandoned without clear value, guardrails, and change management, and the majority of firms have struggled to scale AI beyond pilots (Bain & Company, 2025). The strategic implication is that the defensibility which matters is not demonstrated capability but realized, governed, trusted value in production, which is a far higher bar and one that the surrounding infrastructure, rather than the model, determines.

The synthesis that closes both this module and the series is that AI strategy is the discipline of building durable advantage in an environment where the core capability is rented and commoditizing, which returns the product leader to the classical frameworks rather than away from them. The agenda is to pursue AI-native reconceptions rather than AI-enabled additions, to position deliberately within the shift toward agentic execution rather than be substituted by it, to internalize that the model is an engine and never a moat, and to build defensibility in the layers that survive commoditization, namely proprietary data loops, workflow ownership and embedding, network effects, and trust infrastructure. The product leaders who will define the next decade are not those with access to the best model, since that access is becoming universal, but those who understand that in an age of rented intelligence, durable advantage is built in everything that surrounds the intelligence, which is exactly the strategic work the preceding nineteen modules were preparing them to do.

References

Andreessen Horowitz. (2024). Who owns the generative AI platform? https://a16z.com/who-owns-the-generative-ai-platform/

Bain & Company. (2025). Will agentic AI disrupt SaaS? https://www.bain.com/insights/will-agentic-ai-disrupt-saas-technology-report-2025/

V7 Labs. (2024). Are data moats dead in the age of AI? https://www.v7labs.com/blog/data-moats-a-guide

Agrawal, A., Gans, J., & Goldfarb, A. (2022). Power and prediction: The disruptive economics of artificial intelligence. Harvard Business Review Press.

Calibrating to the Level of Uncertainty

The first discipline, which Courtney et al. (1997) contribute and which most teams skip, is to diagnose how much uncertainty actually obtains before choosing how to respond, because the appropriate strategic posture differs by level. They distinguish a spectrum running from a future clear enough for a confident forecast, through a future with a small set of discrete possible outcomes, to a future with a bounded range of outcomes, and finally to genuine ambiguity in which even the range cannot be specified. The strategic error is to treat all uncertainty as though it were the first level, applying point forecasts to situations that demand scenarios or options, and the corrective is to match the instrument to the level, reserving confident forecasting for the rare cases that warrant it and deploying scenario and option methods where the future is genuinely plural. Owing to the pace of the AI transition, most product decisions now sit at the third or fourth level, where a bounded range or genuine ambiguity prevails, which is why the instruments that follow have become essential rather than optional.

Assumption Mapping: Exposing What a Strategy Depends On

Every strategy rests on assumptions, and the difference between a robust strategy and a fragile one is largely whether those assumptions have been made explicit and tested. Assumption mapping, developed as a core practice by Bland and Osterwalder (2019), is the discipline of surfacing the beliefs a strategy depends upon and arranging them by two dimensions, namely how important each assumption is to the strategy’s success and how much evidence currently supports it. The assumptions that are simultaneously most important and least evidenced are the ones that should be tested first, because they carry the greatest risk of invalidating the entire strategy and the team currently knows the least about them. The strategic value of this practice is that it redirects effort from building on unexamined beliefs to testing the beliefs on which everything else rests, which is the most efficient possible allocation of learning, since it resolves the largest uncertainty at the lowest cost before significant resources are committed.

The discipline this imposes on a product organization is to treat a strategy not as a plan to execute but as a set of hypotheses to validate, identifying the load-bearing assumptions, designing the cheapest experiments that would disconfirm them, and committing resources in proportion to the evidence accumulated rather than to the confidence asserted. In the context of AI, where strategies routinely rest on assumptions about how model capabilities will advance, how users will trust autonomous systems, and how quickly costs will fall, assumption mapping is the instrument that prevents an organization from building an elaborate strategy on a capability forecast it has never examined.

Scenario Planning: Preparing for Futures Rather Than Predicting One

Where assumption mapping tests the beliefs underlying a single strategy, scenario planning prepares an organization for multiple plausible futures, and its purpose is frequently misunderstood. The objective of scenario planning, as developed by Schoemaker (1995) from the practice pioneered at Shell, is not to predict which future will occur but to construct a small set of distinct, internally coherent narratives about how the future could unfold, so that the organization develops the perceptual range to recognize each as it emerges and the prepared responses to act when it does. The value lies in the process as much as the product, because the discipline of imagining several genuinely different futures breaks the organization’s habit of planning for a single extrapolation of the present, which is the habit that leaves it blindsided when the future diverges from the forecast.

The practice constructs typically three or four scenarios around the most important and most uncertain driving forces, develops each into a coherent story, and then stress-tests the strategy against all of them, asking which choices succeed across multiple futures and which depend on a single future obtaining. The strategic payoff is the identification of robust moves that perform acceptably across scenarios and the early-warning indicators that signal which scenario is materializing, which together convert an unpredictable future from a threat into a set of conditions the organization has already rehearsed. The AI transition is the canonical subject for this method, since its trajectory admits genuinely different futures regarding the pace of capability advance, the structure of the model-provider market, and the regulatory environment, and a product strategy that is robust across these scenarios is far more valuable than one optimized for the single future its authors happen to expect.

Optionality Thinking: Structuring Commitments to Preserve Flexibility

The third instrument reframes how commitments themselves are structured, drawing on the logic of real options and on Taleb’s (2012) argument that some strategies benefit from volatility rather than merely surviving it. Optionality thinking treats an investment not as a binary commitment to a forecast outcome but as the purchase of the right, without the obligation, to pursue an opportunity once uncertainty resolves, which means that under high uncertainty the most valuable moves are frequently those that buy information and preserve the ability to scale up or abandon cheaply rather than those that commit fully in advance. A strategy structured as a sequence of staged options, in which small investments resolve key uncertainties and earn the right to larger investments, dominates a strategy structured as a single large commitment when the future is genuinely plural, because it limits the downside of being wrong while preserving the upside of being right.

Taleb’s (2012) contribution is to distinguish strategies that are merely robust, surviving a range of futures unchanged, from those that are antifragile, gaining from the disorder, and to observe that a portfolio of small bounded-loss bets with unbounded upside is positioned to benefit from precisely the volatility that destroys strategies built on point forecasts. In the context of AI, optionality thinking counsels the product leader to make many small, staged, cheaply reversible bets on emerging capabilities rather than a single large bet on a predicted trajectory, since the trajectory is genuinely uncertain and the value of preserving the ability to pivot as it resolves is high.

The synthesis for the product leader is that strategy under uncertainty is not a weaker form of strategy but a more honest and ultimately more powerful one. The agenda is to first diagnose the level of uncertainty rather than defaulting to confident forecasting, to map and test the load-bearing assumptions on which the strategy depends before committing to it, to prepare for a small set of distinct futures through scenario planning rather than betting on a single extrapolation, and to structure commitments as staged options that preserve flexibility and limit downside while retaining upside. A product organization that adopts these instruments will make fewer confident pronouncements about the future and far better decisions within it, which, in a regime defined by a future that refuses to hold still, is the only durable form of strategic advantage.

References

Bland, D. J., & Osterwalder, A. (2019). Testing business ideas: A field guide for rapid experimentation. Wiley.

Courtney, H., Kirkland, J., & Viguerie, P. (1997). Strategy under uncertainty. Harvard Business Review, 75(6), 67–79. https://hbr.org/1997/11/strategy-under-uncertainty

Schoemaker, P. J. H. (1995). Scenario planning: A tool for strategic thinking. Sloan Management Review, 36(2), 25–40.

Taleb, N. N. (2012). Antifragile: Things that gain from disorder. Random House.

The North Star Metric: One Measure of Created Value

The north star metric, a framework associated with John Cutler and developed extensively at Amplitude, is the single measure that best expresses the value a product delivers to its customers, chosen so that the entire organization can align its work behind moving it (Amplitude, n.d.). The discipline of the north star is not that it reduces measurement to one number but that it forces an organization to articulate, in one quantity, what value it actually creates, which is a harder and more clarifying exercise than most teams expect. A well-chosen north star measures customer value rather than company extraction, which is why a measure such as the volume of meaningful work a product helps customers complete is a better north star than the revenue that value eventually produces, since the former is a cause the team can influence and the latter is an effect that arrives too late to steer.

Cutler offers a counterintuitive but essential criterion, which is that if a team can move its north star directly, it is probably not a good north star, because the purpose of the measure is to sit one level beyond direct manipulation so that it provokes the organization to reason about why it moves rather than to game it (Amplitude, n.d.). A north star that can be inflated by a single team’s tactical action is a metric the organization will inflate, whereas a north star that can only be moved by genuinely creating more customer value is a metric whose pursuit aligns the organization with its customers. The strategic function of the north star, therefore, is less measurement than alignment, providing a shared definition of value that coordinates otherwise divergent teams.

Leading and Lagging Indicators: The Architecture of Foresight

The deepest structural distinction in strategic measurement is between leading and lagging indicators, and the failure to observe it is the most common reason organizations learn the truth too late to act on it. A lagging indicator reports an outcome that has already occurred, such as revenue, retention realized over a past period, or churn already incurred, and its defining property is that by the time it moves, the events that determined it are complete and beyond influence. A leading indicator measures the behaviors and conditions that precede and predict the outcome, and its defining property is that it moves early enough that the organization can still act on what it foretells. The strategic discipline is to manage primarily by leading indicators while validating by lagging ones, since steering by lagging indicators alone is, as the metaphor goes, navigating by the wake.

The relationship between the two should be made explicit as a causal chain rather than left implicit, since a leading indicator earns its status only if there is a credible mechanism by which it produces the lagging outcome. The most rigorous measurement systems articulate the chain from the input behaviors a team can influence, through the leading indicators those behaviors move, to the lagging outcomes the business ultimately cares about, so that the organization understands not merely what it is measuring but why each measure should predict the next. A north star metric is typically positioned as a leading indicator of the lagging financial outcomes, which is precisely why revenue, a lagging indicator, makes a poor north star despite its obvious importance (Amplitude, n.d.).

The Revenue and Engagement Traps

Two specific traps corrupt strategic measurement so reliably that they deserve to be named. The first is the revenue trap, the temptation to elevate revenue or its proximate cousins to the central steering metric, which fails not because revenue is unimportant but because it is a lagging indicator on which, as Cutler observes, what is done is done, meaning that by the time revenue reflects a problem the opportunity to influence it has passed (Amplitude, n.d.). An organization that steers by revenue is perpetually reacting to outcomes it can no longer change, whereas an organization that steers by the leading indicators of customer value is acting while action still matters.

The second is the engagement trap, the elevation of engagement metrics such as time spent, sessions, or clicks to the status of success measures, which fails because engagement is not intrinsically valuable and is often inversely related to the value the customer actually seeks. A productivity tool whose users spend more time in it may be delivering more value or may be confusing them, and the raw engagement number cannot distinguish the two, which is what makes engagement a quintessential vanity metric, one that rises in a way that flatters the team while inspiring no action and predicting no durable outcome. Ries (2011) framed the underlying distinction as the difference between vanity metrics that look impressive and actionable metrics that inform decisions, and the diagnostic Cutler offers is precise: if a number rises and the only response is satisfaction, and if it falls and the team does not change its strategy, it is a vanity metric regardless of how prominent it sits on the dashboard.

Strategic Measurement in the Age of AI

The AI transition sharpens every one of these distinctions and introduces a specific new hazard. The hazard is that AI products generate seductive engagement and adoption signals, since novelty drives high initial usage that registers as success on exactly the engagement metrics most prone to the vanity trap, which means an AI feature can appear to be winning on the dashboard while retaining no one. The corrective is to insist that the north star measure realized customer value rather than interaction volume, and to weight the leading indicators of durable value, such as whether users return to accomplish meaningful work after the novelty fades, over the engagement spike that novelty produces. The transition also raises the importance of measuring outcome quality directly, because an AI product’s value depends on whether its outputs are correct and trusted, which is a dimension that conventional engagement and revenue instruments do not capture and which must be measured deliberately if the team is to know whether the product is genuinely working.

The synthesis for the product leader is that strategic measurement is the deliberate construction of a system that measures what compounds rather than what comforts. The agenda is to anchor the organization on a north star that expresses created customer value and sits one level beyond direct manipulation, to architect the measurement system as an explicit causal chain from influenceable inputs through leading indicators to lagging outcomes, to refuse the revenue trap of steering by a lagging financial measure and the engagement trap of celebrating interaction volume as though it were value, and to recognize that the AI transition makes vanity signals more seductive and outcome quality more essential to measure. A product organization that measures this way will sometimes report less flattering numbers than one optimizing for vanity, and it will be the one that knows, early enough to act, whether it is actually winning.

References

Amplitude. (n.d.). The North Star playbook: The guide to discovering your product’s North Star [J. Cutler]. https://amplitude.com/north-star

McClure, D. (2007). Startup metrics for pirates (AARRR).

https://500hats.typepad.com

Ries, E. (2011). The lean startup: How today’s entrepreneurs use continuous innovation to create radically successful businesses. Crown Business.

The Three Horizons: Allocating Across Time

The foundational portfolio framework is the three horizons model articulated by Baghai, Coley, and White (1999), which distinguishes investments by their temporal distance from current returns. The first horizon comprises the core business that generates current revenue and demands continual defense and optimization; the second horizon comprises emerging opportunities that are not yet profitable but are on a credible path to becoming the next core; and the third horizon comprises options on genuinely new businesses whose viability is uncertain and whose returns, if any, lie far in the future. The value of the framework is that it makes the temporal balance of the portfolio visible, and its central warning is that organizations naturally over-invest in the first horizon because its returns are immediate and measurable, while starving the second and third horizons whose returns are deferred and uncertain, thereby mortgaging the future to optimize the present.

The discipline the framework imposes is to fund all three horizons concurrently rather than sequentially, since an organization that waits until the core declines before investing in the next horizon will find that the second-horizon business it needed required years of cultivation it failed to begin. The horizons are not stages a company passes through but parallel investments it must sustain simultaneously, and the product leader’s task is to ensure that the portfolio carries live bets in all three at all times, with the recognition that the horizons themselves are compressing as technological change accelerates.

Core, Adjacent, and Transformational: Allocating Across Novelty

Where the three horizons organize the portfolio by time, the innovation ambition matrix introduced by Nagji and Tuff (2012) organizes it by novelty, distinguishing core initiatives that optimize existing products for existing customers, adjacent initiatives that extend the business into new markets or new capabilities, and transformational initiatives that create new offerings for markets that do not yet exist. The two frameworks are complementary lenses on the same portfolio rather than competitors, since novelty and time correlate but are not identical, and reasoning about both prevents the common error of treating a long-dated incremental project as though it were genuinely transformational.

The empirical contribution of Nagji and Tuff (2012) is the most quoted and least heeded finding in portfolio management, which is that across the companies they studied, a balanced allocation of roughly seventy percent of innovation resources to core, twenty percent to adjacent, and ten percent to transformational correlated with superior share-price performance, while, in a striking asymmetry, approximately seventy percent of the long-term return came from the transformational ten percent. The strategic implication of this asymmetry is profound: the small fraction of the portfolio that organizations most readily cut under pressure is the fraction that generates the disproportionate share of long-run value, which means that the discipline of protecting the transformational allocation is not a luxury of well-resourced firms but the mechanism by which durable returns are actually produced. The authors are careful, and the product leader should be too, that the specific ratio is a reference point rather than a law, since the appropriate allocation varies by industry, competitive intensity, and company stage; the discipline is to know the current allocation, to set a deliberate target, and to manage the gap, rather than to discover after the fact that the portfolio drifted entirely into the core.

Innovation Allocation as an Act of Will

The reason allocation must be treated as a deliberate act rather than an emergent outcome is that the organizational forces acting on a portfolio all push in the same direction, toward the core. Core initiatives have clearer business cases, more confident estimates, more vocal internal advocates, and more immediate metrics, which means that in any unmanaged prioritization process they will crowd out adjacent and transformational work by appearing more rational at every individual decision point. Owing to this systematic bias, a portfolio left to optimize itself will converge on the present, which is why the allocation must be set as a budget and defended as a commitment, insulated from the quarter-to-quarter pressure that would otherwise consume it. The product leader who does not ring-fence the transformational allocation will not have one, regardless of stated intentions, because the sum of locally rational decisions will spend it on the core.

Portfolio Thinking in the Age of AI

The AI transition acts on portfolio thinking in two consequential ways. The first is that it compresses the horizons, moving capabilities that would recently have been third-horizon bets, such as autonomous agents performing end-to-end workflows, into the second and even the first horizon within a single planning cycle, which means that the temporal distance the framework assumes is shorter than it has historically been and the cost of neglecting the transformational allocation is incurred sooner. The second is that it raises the stakes of the transformational bet specifically, because AI-native reconceptions of a category tend to be transformational rather than incremental, which places the most important AI opportunities precisely in the portfolio quadrant that organizational gravity most reliably starves. A product organization that runs its AI work entirely as core optimization, adding features to the existing product, will find that the genuinely transformational AI opportunity was located in the quadrant it declined to fund, and that a competitor willing to make the transformational bet has redefined the category.

The synthesis for the product leader is that portfolio thinking is the antidote to the local optimization that prioritization alone cannot prevent. The agenda is to maintain live bets across all three horizons concurrently rather than sequentially, to allocate deliberately across core, adjacent, and transformational novelty with explicit awareness that the transformational fraction generates the disproportionate long-run return, to defend the transformational allocation as a ring-fenced budget against the organizational gravity that would consume it, and to recognize that the AI transition has both compressed the horizons and concentrated the most important opportunities in the transformational quadrant. A portfolio managed this way will feel less efficient quarter to quarter than one optimized entirely for the core, and it will be the one that is still relevant when the regime changes.

References

Baghai, M., Coley, S., & White, D. (1999). The alchemy of growth: Practical insights for building the enduring enterprise. Perseus Books.

Nagji, B., & Tuff, G. (2012). Managing your innovation portfolio. Harvard Business Review, 90(5), 66–74. https://hbr.org/2012/05/managing-your-innovation-portfolio

Christensen, C. M. (1997). The innovator’s dilemma: When new technologies cause great firms to fail. Harvard Business School Press.

Cost of Delay: The Quantity That Should Govern Sequencing

The single most important concept in economic prioritization is cost of delay, which Don Reinertsen (2009) elevated to the central place in product economics with the instruction that if a team quantifies only one thing, it should quantify the cost of delay. Cost of delay is the value foregone for each unit of time that a valuable outcome is postponed, and its power is that it converts the abstract question of importance into a concrete economic quantity. When a capability would generate a given value per month once shipped, every month of delay forfeits that value irretrievably, which means that the true cost of a feature includes not only the resources to build it but the accumulated value lost while it remained unbuilt. The reason this matters strategically is that teams routinely optimize for the size of the prize while ignoring the meter that runs against them, and a portfolio sequenced without regard to cost of delay will systematically deliver value later than necessary even when every individual decision looked reasonable.

Reinertsen’s (2009) operational contribution is to combine cost of delay with the duration of work into a sequencing rule, weighted shortest job first, which orders work by dividing the cost of delay by the job size so that items with the highest economic urgency per unit of effort are delivered first. The Scaled Agile Framework later formalized the cost-of-delay numerator as a sum of business value, time criticality, and risk reduction or opportunity enablement, divided by job size, which gives teams a tractable means of estimating relative urgency without spurious precision (Scaled Agile, n.d.). The strategic value of this rule is that it makes the meter visible: it forces the team to ask not merely what is most valuable but what is most valuable per unit of time it consumes, which is the only question whose answer maximizes the value delivered across a constrained schedule.

Opportunity Cost: The Discipline of Reasoning About the Road Not Taken

Cost of delay measures the price of postponing a chosen item; opportunity cost measures the price of the items not chosen, and it is the concept that most distinguishes strategic prioritization from backlog management. Every decision to build something is simultaneously a decision not to build everything else the same resources could have produced, and the true cost of any initiative is the value of the most valuable alternative it displaces. The reason this is so easily neglected is that the displaced alternative is invisible: it generates no failure to point to, no metric that declines, only a value that was never created and therefore never missed. The strategic discipline is to make the invisible alternative explicit by requiring, for any significant commitment, an articulation of what the team is choosing not to do as a consequence, since a commitment whose opportunity cost cannot be named is a commitment that has not been genuinely reasoned about.

In the context of AI, opportunity cost has sharpened considerably, because the menu of credible alternatives has expanded and the relative value of items on it is changing quarter to quarter as model capabilities advance. A capability that was uneconomical to build last year because it required bespoke machine learning may now be assembled atop a foundation model in a fraction of the time, which means the opportunity cost of every other backlog item has risen, since the resources spent elsewhere now displace newly cheap and valuable AI capabilities. The product leader who is not periodically re-pricing the backlog against this shifting menu is, in effect, paying an opportunity cost that compounds silently.

Strategic Bets Versus Incremental Bets

The deepest error in prioritization is to apply a single economic logic to two categories of decision that require different logics, namely incremental bets and strategic bets. Incremental bets are improvements to a known product serving a known market, where the value is estimable, the risk is modest, and the cost-of-delay and weighted-shortest-job-first machinery applies cleanly, because the quantities can be estimated with enough confidence to sequence on. Strategic bets are commitments to a new capability, market, or business model where the value is genuinely uncertain, the time horizon is long, and the principal return is not the immediate value but the option the bet creates and the learning it produces. Subjecting a strategic bet to the same near-term economic scoring as an incremental feature will almost always rank it poorly, because its near-term value is low and uncertain by construction, which is precisely how organizations systematically starve the bets that determine their future while feeling rigorous for doing so.

The strategic discipline, therefore, is to separate the two budgets and to evaluate them on different criteria, scoring incremental bets on economic urgency per unit of effort and evaluating strategic bets on the magnitude and durability of the advantage they could create and the cost of learning whether they will. This separation is the operational link between prioritization and the portfolio thinking developed in the next module, since a backlog that mixes the two categories on one ranked list will reliably under-invest in the transformational work, given that incremental items will always present a more favorable near-term ratio.

The Synthesis for the Product Leader

The agenda this module sets for the product leader is to treat prioritization as applied economics rather than as list management. The first discipline is to quantify cost of delay, even roughly, so that the meter running against every delayed item becomes visible and sequencing reflects value delivered over time rather than value in isolation. The second is to name the opportunity cost of every significant commitment, making the displaced alternative explicit so that the road not taken is reasoned about rather than ignored, with particular attention to how the AI transition has raised the value of the alternatives the backlog displaces. The third is to refuse to evaluate strategic bets and incremental bets on a single ranked list, separating their budgets and their criteria so that the organization does not starve its future to optimize its present. A product leader who institutionalizes these three disciplines will find that prioritization stops being a recurring negotiation over a spreadsheet and becomes what it should be, namely the deliberate economic allocation of the only resource that cannot be replenished.

References

Reinertsen, D. G. (2009). The principles of product development flow: Second generation lean product development. Celeritas Publishing.

Scaled Agile. (n.d.). Weighted shortest job first (WSJF). https://framework.scaledagile.com/wsjf

Christensen, C. M. (1997). The innovator’s dilemma: When new technologies cause great firms to fail. Harvard Business School Press.

APIs as Product, Not Plumbing

The most consequential reframing in this module is that an application programming interface is a product with its own users, value proposition, and growth dynamics, rather than a technical afterthought. The clearest demonstration of this principle is the mandate Jeff Bezos is reported to have issued at Amazon, requiring that all teams expose their data and functionality exclusively through service interfaces, that these interfaces be designed to be externalizable as if outside developers would consume them, and that no team communicate through any other path. The strategic effect of this discipline was profound: by forcing every internal capability to be expressed as a clean, externalizable service, Amazon created the architectural precondition for externalizing its computing infrastructure entirely, first internally and then to the world as Amazon Web Services, which became one of the most valuable businesses in technology (Nordic APIs, 2021). The lesson for the product leader is that treating APIs as products, with the same attention to developer experience, documentation, and reliability that one gives a user-facing surface, converts internal capability into a platform for external value creation.

The mechanism by which an API creates leverage is that it lets parties the product does not employ build value on the product’s behalf, which means the rate of value creation is no longer bounded by the product team’s own capacity. The metric that captures this leverage is the number of consumers building on the interface relative to the effort of maintaining it, since a well-designed API that many developers depend upon generates compounding value and switching costs that a closed product cannot match.

Partnerships and the Logic of Complementarity

Partnerships are the second lever, and their strategic logic is complementarity rather than mere co-marketing. A partnership creates durable value when each party supplies something the other cannot easily build and when the combination is worth more to the customer than either alone, which is a more demanding test than the logo-exchange partnerships that fill many slide decks. The product leader’s discipline is to evaluate a prospective partnership by asking what asset the partner controls that the product genuinely cannot replicate, whether the partner’s incentives are durably aligned with the product’s, and whether the integration deepens over time into mutual dependence or remains a shallow connection either party can sever without cost. The partnerships that matter strategically are those that embed the product into a workflow or distribution channel the partner controls, because that embedding becomes a switching cost for the customer and a barrier for the rivalry.

Distribution Leverage: The Force Multiplier PMs Underuse

Distribution leverage is the lever product leaders most consistently underweight, owing to a cultural conviction that a sufficiently good product distributes itself, which is true far less often than builders wish. The strategic reality is that the channel through which a product reaches users is frequently more decisive than the product’s marginal quality advantage, because a product that is slightly worse but embedded in a channel the customer already inhabits will defeat a product that is slightly better but requires the customer to discover and adopt it independently. Ecosystem strategy treats distribution as a deliberate design problem: integrating into the marketplaces, app directories, and platforms where the target customer already transacts converts the platform’s installed base into the product’s distribution, which is leverage no advertising budget can match. The product leaders who understand this seek out the channels their customers already trust and design the product to be discoverable and installable within them, rather than building in isolation and hoping for unaided demand.

Platform Expansion and the Sequencing of Ambition

Platform expansion is the fourth lever and the most ambitious, describing the move from a product that serves a use case to a platform that hosts an ecosystem of use cases built by others. The sequencing of this ambition is where most attempts fail, because a product that opens itself to third-party developers before it has achieved sufficient value and scale on its own offers those developers no audience worth building for, which means the platform’s network effect cannot ignite. The durable pattern, visible across the products that successfully became platforms, is to first win decisively as a product, accumulating the user base that makes the platform attractive, and only then to open the surfaces that invite external builders, so that the platform launches into demand rather than into emptiness. Premature platform ambition is a recurring and expensive error, and the product leader’s discipline is to recognize that platform expansion is earned by product success rather than substituted for it.

Ecosystem Strategy in the Age of AI

The AI transition raises the strategic weight of ecosystem strategy because it simultaneously commoditizes capability and reshapes distribution. As foundation models make raw capability broadly available, the differentiation that survives shifts toward integration depth, namely how deeply a product is embedded in the systems of record, workflows, and data sources where work actually happens, since this embedding is precisely what a competitor renting the same model cannot quickly replicate. At the same time, a new and consequential distribution surface is emerging in the form of agentic systems that act on a user’s behalf, which raises the prospect that the relevant ecosystem to integrate into is no longer only the human-facing application directory but the set of protocols and interfaces through which autonomous agents discover and invoke capabilities. The product leader who anticipates this is asking not only how human users will find the product but how an agent will, and is designing the product’s interfaces to be invokable by autonomous systems, which is the contemporary equivalent of designing for the API economy a decade ago.

The synthesis is that ecosystem strategy has moved from a peripheral concern to a central determinant of defensibility, precisely because the AI transition has eroded the feature-level differentiation on which many products previously relied. The product leader’s agenda is therefore to treat APIs as products, to pursue partnerships on the demanding test of genuine complementarity, to design distribution as deliberately as one designs features, to sequence platform expansion so that it launches into earned demand, and to extend all of this to a world in which agents as well as humans are the users a product must reach. In an era when capability is rented, the ecosystem is the product, and the leaders who internalize this will compete on a dimension their feature-focused rivals do not yet recognize as the battleground.

References

Cusumano, M. A., Gawer, A., & Yoffie, D. B. (2019). The business of platforms: Strategy in the age of digital competition, innovation, and power. Harper Business.

Nordic APIs. (2021). The Bezos API mandate: Amazon’s manifesto for externalization. https://nordicapis.com/the-bezos-api-mandate-amazons-manifesto-for-externalization/

Parker, G. G., Van Alstyne, M. W., & Choudary, S. P. (2016). Platform revolution: How networked markets are transforming the economy and how to make them work for you. W. W. Norton.

A Network Effect Is Not One Thing

The foundational correction is definitional: a network effect exists when each additional user makes the product more valuable to other users, and the strength of the effect, which is what determines defensibility, varies enormously by type. NFX’s taxonomy is useful precisely because it ranks these effects by durability, observing that physical and protocol-based effects tend to be the most defensible because they combine the direct network effect with additional barriers such as capital intensity and embedding, while social and personal-utility effects, though real, are more contestable (Currier, 2018). The product leader’s first task is therefore diagnostic, since the design implications of a marketplace effect, in which buyers attract sellers and sellers attract buyers, differ fundamentally from those of a data network effect, in which usage improves a shared model that benefits all users, which differ again from a personal-utility effect, in which the product becomes valuable because the people one personally needs to reach are present.

Currier (2018) further argues that network effects are best understood as one of a small set of structural defensibilities, alongside brand, embedding, and economies of scale, and that the strongest businesses stack several together so that each reinforces the others. This stacking insight matters for product strategy because it reframes the goal from possessing a single network effect to architecting a system in which a network effect, switching costs, and scale advantages compound, which is far harder for a competitor to assail than any one of them alone.

Marketplace Dynamics and the Cold Start Problem

Marketplaces are the canonical two-sided network effect, and their dynamics expose the central difficulty of all network-effect businesses, which is that the same mechanism that makes them defensible at scale makes them nearly worthless at the start. A marketplace with no sellers offers nothing to buyers and a marketplace with no buyers offers nothing to sellers, so the network effect runs in reverse before it runs forward. Andrew Chen (2021) frames this as the cold start problem and argues that it is solved not by launching the whole network but by igniting a small, dense, self-sustaining cell of the network, an atomic network, in which the effect is locally positive even though the global network is empty. The historical pattern confirms the logic: marketplaces and social products that endured typically seeded a narrow segment first, such as a single city, a single campus, or a single category, and only generalized once the atomic network could stand on its own.

The strategic implication for the product leader is that network-effect businesses require a different sequencing discipline than feature businesses, because the early roadmap must prioritize achieving critical density within a deliberately narrow segment over breadth, and metrics must be read per atomic network rather than in aggregate, since healthy local density can be masked by sparse global numbers and sparse local density can be flattered by broad but disconnected adoption.

Platform Strategy and Defensibility

A platform extends the network-effect logic by opening the product to third parties who build on top of it, converting a product into an ecosystem in which the platform’s value grows with the contributions of complementors rather than only with its own development. Parker, Van Alstyne, and Choudary (2016) characterize the resulting shift as a move from controlling a linear value chain, in which a firm creates value and ships it downstream, to orchestrating an ecosystem, in which the firm’s principal task is to facilitate value-creating interactions among participants it does not own. The defensibility of a platform comes from the combination of the network effect among its participants and the accumulated investment those participants have made in building upon it, which together raise switching costs to a level no feature comparison can overcome. The enduring examples, in which independent developers, complementary hardware, and a large installed base mutually reinforce one another, demonstrate that platform defensibility is a property of the ecosystem’s structure rather than of any single product decision.

Network Effects and Platforms in the Age of AI

The AI transition acts on network effects in two opposing directions that a product leader must reason about carefully. In the first direction, AI threatens certain network effects by lowering the value of aggregated content and the friction of switching, since a sufficiently capable model can synthesize what a content network previously made uniquely available, which weakens the personal-utility and content-aggregation effects that some products relied upon. In the second direction, AI introduces and strengthens the data network effect, in which accumulated usage data trains a model that improves the product for all users, and this effect is particularly powerful because, unlike a rented foundation model, the proprietary data and the specialized model it produces are assets a competitor cannot acquire by renting the same base capability. The most defensible AI products will be those that convert their network into a data network effect, so that scale produces a model advantage that widens with each additional user, while the least defensible will be those whose only network is one that a general-purpose model can now route around.

The synthesis for the product leader is that network effects remain the strongest available defensibility in the digital economy, but that the AI transition has reordered which types are durable and has raised the premium on stacking. A product whose advantage rests on a single contestable network effect is more exposed than its leaders typically believe, because AI is precisely the force that contests it; a product that stacks a network effect with a proprietary data loop, switching costs from embedding, and scale economics builds an architecture that the rivalry, even when armed with the same models, struggles to replicate. The discipline, therefore, is to treat network effects not as a fortunate property the product happens to have but as an architecture the product leader is responsible for designing, diagnosing by type, igniting locally, and reinforcing by stacking, with full attention to how the AI transition strengthens some layers of that architecture and erodes others.

References

Chen, A. (2021). The cold start problem: How to start and scale network effects. Harper Business.

Currier, J. (2018). The network effects manual: 16 different network effects (and counting). NFX. https://www.nfx.com/post/network-effects-manual

Parker, G. G., Van Alstyne, M. W., & Choudary, S. P. (2016). Platform revolution: How networked markets are transforming the economy and how to make them work for you. W. W. Norton.

The Structural Defect of the Funnel

The funnel’s defect is not that it is wrong about conversion; it is that it is silent about reinvestment. A funnel is, in mathematical terms, additive: if marketing delivers a constant number of new prospects each month and conversion rates hold, the business grows linearly, and growth stalls the moment the input spend stalls (Reforge, 2017). This produces a recognizable organizational pathology in which growth becomes synonymous with budget, every increment of growth requires a proportional increment of spend, and the team is perpetually buying its next cohort rather than earning it. The funnel also fragments the organization, because it encourages separate teams to optimize separate stages, each improving its local conversion rate while no one owns the question of how the whole system perpetuates itself.

A growth loop reframes the unit of analysis from the stage to the cycle. Rather than asking how a prospect moves down a funnel, the loop asks how one cohort of users produces the next cohort of users, and it closes the system by specifying how the output of one cycle is reinvested as the input of the next (Reforge, 2017). The difference is the difference between additive and compounding: when each user reliably generates a fraction of a new user, growth is geometric rather than linear, and the system gains momentum from its own scale rather than depleting a budget. The strategic consequence is that loops, because they integrate product, channel, and monetization into a single self-reinforcing system specific to one company, are markedly harder for competitors to replicate than any single funnel optimization.

Why Loops Matter Strategically, Not Just Tactically

The reason this distinction belongs in a strategy module rather than a growth-tactics playbook is that the choice of loop determines the shape of the entire business, not merely the efficiency of its marketing. A product whose dominant loop is content-driven, in which users generate content that ranks in search and attracts new users who generate more content, will build different teams, accumulate different assets, and defend itself differently than a product whose dominant loop is viral, in which users invite other users directly, or one whose dominant loop is paid, in which revenue from current users funds acquisition of the next. Pinterest and Quora grew principally through content loops in which user-generated pages compounded organic discovery; Dropbox grew through a viral loop in which the act of sharing files recruited the recipients; PayPal seeded a referral loop with direct financial incentives that turned each user into an acquisition channel. The strategic act is to identify which loop the product’s value creation naturally produces and to invest in deepening it, rather than attempting to operate every loop weakly.

Virality as a Property of the Loop, Not an Add-On

Virality is frequently misunderstood as a feature to be bolted on, a referral widget added late to a product whose core offers no native reason to share. The loop framing corrects this by treating virality as a structural property: a viral loop compounds only when the product’s core action inherently exposes or recruits new users, such that growth is a byproduct of use rather than a separate behavior the user must be cajoled into. The viral coefficient, the average number of new users each existing user generates, and the cycle time over which they generate them, jointly determine whether the loop compounds or decays, and the durable cases are those where the sharing is intrinsic to deriving value, as collaboration tools demonstrate when inviting a collaborator is simply how the work gets done. Bolted-on referral mechanics rarely compound because they sit outside the value-creating action and therefore decay as soon as the incentive is withdrawn.

Compounding Systems and the AI Transition

The AI transition reshapes growth loops in a way that makes the most powerful loop available to a new and larger class of products, namely the data-and-model loop. In this loop, usage generates data, the data improves the model, the improved model makes the product more valuable, the increased value attracts more usage, and the cycle compounds, with the distinctive feature that the reinvested output is product quality itself rather than only audience or revenue. This is the loop that confers the most durable advantage in AI products, because the accumulated, usage-derived data is the asset competitors renting the same foundation model cannot replicate. A product leader building anything AI-native should ask explicitly whether the product’s usage feeds a model improvement that no competitor can observe, because in its absence the product is renting capability that everyone can rent, and in its presence the product is building the one loop the rivalry cannot copy.

The same transition, however, introduces a sobering counterforce on the distribution side of the loop, which is that the channels many loops depend upon are themselves being reshaped by AI. Loops that compounded through search-indexed user-generated content now contend with AI-generated answers that satisfy the query without delivering the click, which compresses the organic distribution that powered the content loop for a decade. The strategic implication is not that content loops are dead but that the loop must be re-examined against the channel it actually runs on, since a loop whose distribution mechanism is being intermediated by a model will decay even when the product itself is excellent.

The practical agenda, then, is to stop drawing funnels as though they were strategy and to require, for any significant growth investment, an explicit statement of the loop it feeds: what the user does, what that action produces, how the product reinvests that output, and over what cycle time the reinvestment returns as new growth. A funnel optimization that does not strengthen a loop buys a cohort; a loop investment compounds. In a regime where capability is increasingly rented and rivalry is increasingly fast, the compounding system is the closest thing a product has to a structural advantage, and the discipline of building loops rather than funnels is how that advantage is deliberately, rather than accidentally, created.

References

Andreessen Horowitz. (2024). Who owns the generative AI platform? https://a16z.com/who-owns-the-generative-ai-platform/

Chen, A. (2021). The cold start problem: How to start and scale network effects. Harper Business.

Reforge. (2017). Growth loops are the new funnels [B. Balfour, C. Winters, K. Kwok, & A. Chen]. https://www.reforge.com/blog/growth-loops