I've Seen the Light. It's Not PMF.
Startups lose to organizational entropy before competitors.
There's a moment every founder recognizes. The demos go well. A few customers convert. The retention curve looks acceptable. Investors nod along in pitch meetings. And then, quietly, nothing compounds. Growth plateaus at a number too small to be rounding error and too large to be ignored. That purgatory has a name and everyone calls it the absence of PMF.
But I've come to believe that framing is wrong. Not subtly wrong. Fundamentally wrong. Product-Market Fit - the way founders invoke it, the way blog posts teach it, is almost always a post-hoc rationalization for a phenomenon far more complex and far less romantic than a product slotting neatly into a market. After watching dozens of companies try to find it, and a handful actually achieve it, I'm convinced that PMF is not a destination. It is an equilibrium and it is perpetually at risk of decay.
The real insight isn't how to find PMF. It's understanding all the systems that silently undermine it before you ever get there.
The equation no one writes on a whiteboard
Marc Andreessen's definition — being in a good market with a product that can satisfy that market — is correct the way Newton's laws are correct: useful for most cases, insufficient for the hard ones. The problem is that founders hear it and immediately collapse a multi-variable system into a single axis. They optimize the product. Features, latency, UI, model quality, infrastructure. And they wonder why it doesn't compound.
The fuller picture looks more like this:
A zero in any variable collapses the entire product. This is multiplication, not addition.
This isn't just a rhetorical expansion. The multiplication is intentional. A product with a perfect score on four of five dimensions and a near-zero on one will approach zero overall. In my previous firm, I've watched the company with genuinely transformative technology fail because timing was wrong. I've watched excellent technology fail because the distribution vector didn't match the buying behavior of the customer. These aren't edge cases.
And beneath even this equation lies a more granular one - It's the Probability of Adoption:
If any variable approaches zero, adoption collapses — regardless of product quality.
The formula reveals how many silent ways a startup can die. A product solving a painful problem with no identifiable budget owner. A workflow improvement that's useful but not daily. A tool that the end user loves but that requires a procurement decision four levels above them. Each configuration is a plausible-sounding startup that quietly runs out of road.
Problem 01 — The "cool product, non-urgent problem" trap
The most deceptive signal in all of early-stage building is the enthusiastic demo. Users say "wow." They say "I've needed this forever." They ask thoughtful questions. They sign up for the beta. And then they don't change their behavior at all.
What's happening is a cognitive phenomenon that founders persistently underestimate: humans adapt to inefficiency with remarkable grace. In consumer contexts, this manifests as habituation — people stop seeing the friction they live inside. In B2B contexts, something more structural occurs. Organizational pain diffuses. No single person feels enough acute discomfort to absorb the switching cost, the retraining cost, the social cost of being the person who pushed for a change that might not work. The pain is real, it's just distributed across enough people that no individual threshold is crossed.
Admiration is not urgency. These are completely different psychological states that happen to produce similar surface behaviors in a pitch or a demo. Urgency triggers action. Admiration triggers compliments. The founder who cannot distinguish between them will spend eighteen months building a product for people who love it but won't pay for it.
The market doesn't reward impressive technology. It rewards the removal of pain that is acute, frequent, and owned by someone with a budget.
Problem 02 — Being early feels identical to being wrong
There is a particular kind of torture reserved for founders who are correct about the future but incorrect about when the future arrives. I have been in that position. It is not instructive in the moment. It simply feels like failure.
Markets evolve through at least four distinct mechanisms simultaneously: infrastructure maturity, cost curves, behavior shifts, and cultural normalization. A product can be technically sound, economically viable in theory, and behaviorally sensible — and still fail because the culture hasn't metabolized the category yet. Remote collaboration tools pitched in 2016. AI coding assistants in 2019. Async video in 2015. Each of these eventually became enormous. The companies that built them at the wrong moment mostly didn't survive to participate in that growth.
The tragic dimension of timing errors is that they're nearly impossible to diagnose from the inside. The feedback you receive — low conversion, high churn, shallow engagement — is identical to the feedback you'd receive if your product were simply bad. There's no error message that reads "correct idea, wrong year." You have to develop a kind of second-order market awareness: not just "is this a real problem?" but "has the infrastructure, psychology, and workflow context required for this solution to be adopted actually arrived yet?" Most founders skip that question entirely.
Problem 03 — Users want the outcome, not the workflow change
This is the subtlest and most persistent failure mode in B2B SaaS, and it has become dramatically more acute in the AI era. Customers want better clips. Faster output. Higher reach. Lower cost. More data. They want all the outputs that would accrue from adopting your product — and they want them without retraining their team, restructuring their SOPs, adding a new QA layer, or absorbing the uncertainty that comes with any behavioral change at scale.
This creates what I've come to think of as innovation resistance — a force that grows proportionally with the depth of workflow integration your product requires. The deeper the change, the harder adoption becomes, regardless of the value created. A tool that sits on the edge of a workflow, augmenting one step without touching anything upstream or downstream, has a fundamentally easier adoption path than a tool that restructures the workflow itself. Both might create equivalent economic value. The adoption curves look nothing alike.
The implication for product strategy is significant: the product that wins is often not the one that does the most, but the one that inserts itself most invisibly into existing motion. The wedge that doesn't feel like a wedge. The tool that makes the current workflow work slightly better before it eventually becomes the new workflow entirely.
Problem 04 — Organizational fit is not the same as user delight
Consumer PMF is, in a meaningful sense, simple. One person encounters your product, decides they like it, and adopts it. The decision tree has one node. B2B PMF is an entirely different kind of problem
Consider what actually has to be true for an enterprise product to achieve real PMF. The end user has to find it valuable. Their manager has to see throughput improvement. Finance has to see a defensible line item. Security has to clear it. IT has to integrate it. Procurement has to process it. Leadership has to sanction the budget. Each of these stakeholders has a different definition of value, a different threshold for action, and different information about what your product actually does.
This means that PMF in B2B is an organizational achievement. The product has to be good enough to win the end user. But the company its messaging, its pricing, its sales motion, its security posture, its support model has to be designed for the full organizational topology of the buyer. Founders who optimize only for the end user frequently build products that are beloved by the people with the least purchasing authority in the organization.
Problem 05 — Usage and dependency are not the same thing
This distinction has ended more startups than people realize. There is an entire class of product engagement that looks like product-market fit but is structurally hollow — what I think of as curiosity adoption. Users try the product. They use it for a few weeks. Retention numbers look reasonable. The cohort curves flatten at a seemingly acceptable level. And then, gradually, usage drifts. The product wasn't integrated into anything. It was sampled.
Real PMF has a different texture. It feels like operational dependency. The test is not whether users return it is whether removal would break something. Whether workflows were restructured around the product. Whether data accumulated inside it that can't easily migrate out. Whether team communication references it. Whether downstream systems integrated with it. When removal becomes genuinely painful not inconvenient, painful - you've crossed into real PMF territory.
Not "users return" — but "users have rebuilt their workflow around you."
The difference between a product people use and a product people depend on is the difference between a feature and an infrastructure. The ones that do rarely achieved it by building more features — they achieved it by making themselves load-bearing.
Problem 06 — AI products compress their own perceived value
This is the problem that is almost unique to the current moment, and it's one I find genuinely fascinating from a behavioral economics perspective. AI products suffer from a counterintuitive pricing paradox: the more automated and seamless the product becomes, the less valuable it appears to the buyer.
Human beings have a deeply ingrained association between effort and value. We trust things that look hard. We're suspicious of things that look easy. When a process that used to take a professional eight hours gets compressed to thirty seconds by an AI model, the buyer's instinct is not "this is worth more." It's "this is now a commodity." The value isn't denied it's perceived as ambient, automatic, and therefore inexpensive. Better AI can reduce pricing power. That is a real and underappreciated risk in every AI product category right now.
The mitigation is to anchor pricing to outcomes and business impact, not to the labor being replaced. The question is never "how much time does this save?" The question is "how much is the outcome worth?" Those numbers are often an order of magnitude apart but getting buyers to think in terms of the latter requires deliberate framing, case studies, and in many cases a fundamentally different sales motion.
Problem 07 — Distribution mismatch is a silent killer
Many products fail not because the product is wrong, but because the channel through which it reaches the market is wrong. The acquisition strategy doesn't match the buying behavior of the customer, and the result is a product that is technically available to the right people but never actually reaches them through the paths they use to discover and evaluate tools.
An enterprise workflow product sold via PLG motion is one version of this failure. The product requires champions, procurement, and multi-stakeholder alignment — but the go-to-market assumes individual sign-up and bottom-up expansion. The individual users who sign up never have the authority to expand. Another version: a low ACV product sold through outbound-heavy enterprise sales. The economics don't work. The cost of acquisition exceeds the lifetime value before you've even built the relationship.
PMF, properly understood, includes the discovery of what I call the natural distribution vector — the channel through which the product reaches buyers in a way that matches their actual behavior and decision-making process. It shapes pricing, positioning, feature priorities, and the entire onboarding experience. Finding the wrong distribution vector can make even a genuinely great product look like it has no market.
Problem 08 — Multi-persona dynamics create unstable PMF
Every B2B product operates within a web of personas — users, buyers, approvers, beneficiaries, and people whose workflows are disrupted by adoption. This web is rarely discussed in PMF conversations, but it is frequently the reason products plateau.
Consider a podcast production tool. The editor cares about quality control. The producer cares about throughput. The founder cares about margins. The creator cares about brand fidelity. These are not all compatible incentives. A product that meaningfully improves one persona's life may simultaneously threaten another's. The editor who takes pride in the craft of clipping may experience an AI clipping tool not as helpful, but as a threat to their professional identity and job security. That resistance doesn't show up in a product review. It shows up as slow adoption, low enthusiasm, and a champion who doesn't champion.
Stable PMF requires not just satisfying the most important persona, but understanding the full ecosystem of humans who will be affected by the product — and designing a value proposition and change management approach that doesn't create internal adversaries in the process of helping the primary buyer.
Problem 09 — Solving the wrong layer of the problem
The deepest product-strategy error, and the one most likely to generate the "cool tool, not mission-critical" response, is optimizing at the wrong layer of the customer's problem stack. Founders tend to solve the surface inefficiency — the thing that's most visible, most technically interesting, most tractable. Buyers care about the business bottleneck — the thing that is actually limiting their growth, their revenue, their capacity.
These are rarely the same thing. You might build a dramatically better clipping tool for a podcast studio. But if the studio's bottleneck is distribution reach, or sponsorship conversion, or audience growth — then improving clipping by 10x barely moves the needle on their actual business outcome. The tool is good. It just doesn't solve the constraint. And products that don't address the constraint don't become mission-critical, regardless of how well they address everything else.
The most valuable products solve constraint bottlenecks. Not workflow inefficiencies. There is a profound and underappreciated difference.
The discipline this requires is the hardest thing in product strategy: looking past the problem the customer is describing to understand the problem behind the problem. The stated need is usually a symptom. The constraint is usually structural, economic, or organizational — and often invisible to the customer themselves until you name it for them.
Problem 10 — Founder-led artificial PMF
This is the failure mode that looks most like success. Early customers stay. Engagement looks real. The numbers aren't embarrassing. And the reason all of this is true is that the founder is personally holding the product together — patching workflows, customizing onboarding, absorbing edge cases, being the support team and the product team and the customer success team simultaneously.
What's being created is a simulation of PMF. The product works because there's a brilliant, committed human manually compensating for everything it doesn't do. Remove that human, and the product falls apart. The customers who seemed retained will churn. The workflows that seemed integrated will revert. The PMF that seemed real will vanish.
Real PMF appears only when the product functions without founder heroics — when a new customer can come in, onboard, reach value, and integrate the tool into their workflow without a founder on the phone. The test is brutal and clarifying: what happens to retention if you go on a two-week vacation and don't check your phone? The answer tells you more about where you actually are than any metric dashboard.
The deepest truth about PMF
After all of this, I want to offer the frame I've found most clarifying the one that has changed how I think about product strategy at every company I've worked since I first articulated it.
PMF is not when people like your product. It's when your product becomes the path of least resistance. That's a fundamentally different target. People don't choose products because they're optimal. They choose them because they're familiar, predictable, safe, and socially validated. Winning products gradually become easier than the old behavior, safer than the old behavior, and more socially accepted than the old behavior. When all three of those conditions are met simultaneously, adoption begins to compound — not because the product got better, but because the environment around it shifted.
This is why the obsession of elite founders is rarely about features or models. It's about behavior design, onboarding friction, trust architecture, switching costs, organizational compatibility, wedge strategy, distribution loops, and economic alignment. These are the dimensions along which PMF is actually built. Software quality is table stakes. Systems alignment is the game.
Products don't usually lose to competitors. They lose to entropy to the accumulated weight of all the misalignments in the system that no one fully mapped. The market that wasn't ready. The persona that felt threatened. The distribution vector that didn't match. The layer of the problem that wasn't the real constraint. The founder who was holding it all together and eventually couldn't.
The light I've seen isn't PMF. It's the full complexity of what PMF actually requires and the humility that comes from understanding how many ways it can silently fail before you ever know it was missing.