Every investor has a story about the deal they passed on. The company that seemed too early, too niche, too weird — and then became something extraordinary. These stories are painful because they reveal the gap between what we thought we knew and what was actually true. But they are also instructive, because they almost always point to the same underlying failure: we mistook noise for signal.
Every investor has a story about the deal they passed on. The company that seemed too early, too niche, too weird — and then became something extraordinary. These stories are painful because they reveal the gap between what we thought we knew and what was actually true. But they are also instructive, because they almost always point to the same underlying failure: we mistook noise for signal.
The signal-to-noise problem is the central challenge of early-stage investing. At the seed stage, the information available about any given company is sparse, ambiguous, and often actively misleading. Founders are optimistic by nature — they have to be. Markets are uncertain. Traction is early. The metrics that matter most are the ones that do not yet exist.
In this environment, the ability to distinguish genuine signal from the noise that surrounds it is the most important skill an investor can develop. And it is one that almost no one talks about honestly.
The Pattern Recognition Trap
The dominant framework for early-stage deal evaluation is pattern recognition: the idea that experienced investors can identify great companies by comparing them to the great companies they have seen before. There is something to this. Experience does produce genuine insight. The investor who has seen fifty SaaS companies scale from zero to $10M ARR has a calibrated sense of what good looks like at each stage.
But pattern recognition has a dark side. When you are looking for patterns, you tend to find them — whether they are there or not. The cognitive bias literature is unambiguous on this point: humans are extraordinarily good at finding patterns in random data. We see faces in clouds, trends in noise, and inevitability in outcomes that were actually highly contingent.
In venture, this manifests as a systematic bias toward founders who look like previous successful founders, companies that resemble previous successful companies, and markets that feel familiar. The research on this is uncomfortable: studies of VC decision-making have consistently found that demographic similarity between investors and founders is a significant predictor of investment decisions, independent of company quality. This is not malice. It is the pattern recognition instinct misfiring.
The antidote is not to abandon pattern recognition — it is to be explicit about what patterns you are actually looking for, and to distinguish between the patterns that are genuinely predictive and the ones that are merely familiar.
The Signals That Actually Matter
After years of evaluating companies, I have come to believe that there are a small number of signals that are genuinely predictive of early-stage success — and a much larger number that feel predictive but are not.
The first genuine signal is founder-market fit: the specific combination of experience, insight, and obsession that makes a particular person the right person to solve a particular problem at a particular time. This is different from general founder quality. A brilliant generalist is not the same as a brilliant specialist. The founder who has lived the problem — who has felt the friction personally, who has tried and failed to solve it with existing tools, who understands the market at a level of specificity that comes only from direct experience — is a categorically different bet from the founder who identified the problem from the outside.
The second genuine signal is early evidence of pull. In the earliest stages of a company, before there is meaningful revenue or growth, the most important question is whether the product is being pulled into the market or pushed into it. Are customers seeking out the product, or is the founder working hard to convince people to try it? Pull is a signal that the product is solving a real problem in a way that people actually value. Push is a signal that the founder believes in the product more than the market does.
The third genuine signal is the quality of thinking under pressure. Every investor meeting is, at some level, a stress test. How does the founder respond to hard questions? Do they get defensive, or do they engage with genuine curiosity? Do they have a clear mental model of their business, or do they rely on slides and rehearsed answers? The founder who can think clearly and honestly under the pressure of an investor meeting is demonstrating a capability that will be essential when the company faces the much greater pressures of building and scaling.
The Noise That Masquerades as Signal
The signals that are not actually predictive are, in many ways, more important to understand — because they are the ones that most reliably lead investors astray.
The most dangerous is the impressive pitch. Founders who are excellent communicators, who can tell a compelling story, who have a polished deck and a confident manner, are systematically overrated by investors. The ability to pitch is a real skill. It is also almost entirely uncorrelated with the ability to build. The best pitches I have ever seen were for companies that failed. The worst pitches I have ever seen were for companies that succeeded. The correlation is not zero, but it is much weaker than most investors assume.
The second dangerous non-signal is the impressive pedigree. Stanford, MIT, Google, Goldman — these are real signals of intelligence and work ethic. They are not signals of entrepreneurial capability. The skills that make someone successful in a structured institutional environment are often the opposite of the skills that make someone successful in the chaos of an early-stage company. I have backed founders with extraordinary credentials who could not execute, and founders with unconventional backgrounds who built extraordinary companies. The pedigree is a prior, not a conclusion.
The third dangerous non-signal is the large market. Every investor knows that market size matters. What fewer investors acknowledge is that market size estimates are almost always wrong — and wrong in a specific direction. The markets that produce the most valuable companies are almost never the markets that looked large at the time of investment. They are the markets that the company itself created or dramatically expanded. Uber did not invest in the taxi market. Airbnb did not invest in the hotel market. The market size at the time of investment is a constraint on the imagination, not a prediction of the outcome.
Building a Better Evaluation Process
The practical implication of all of this is that good deal evaluation requires a deliberate effort to separate the signals that matter from the noise that surrounds them. This means asking different questions than most investors ask.
Instead of "how big is the market?" ask "what would have to be true for this market to be much larger than it currently appears?" Instead of "how impressive is the founder?" ask "what specific experience does this founder have that makes them the right person to solve this specific problem?" Instead of "how good is the pitch?" ask "how does this founder respond when I push back on their assumptions?"
These are harder questions. They require more time and more intellectual honesty. But they are the questions that lead to the investments that matter — the ones where you saw something that the market had not yet seen, and had the conviction to act on it.
The signal is always there. The work is learning to hear it through the noise.
Richard Smullen is the co-founder of Keller Winston and the founder of Pypestream, Inc.
