What counts as noise
Noise is any metric that responds to your effort rather than to the value your product creates. Inbound sign-ups from a press hit. Demo bookings after a conference. Pilot agreements from warm intros. These are real things. They require real effort to produce. But they are not PMF signals because they do not persist without your intervention.
The test is simple: if you stopped doing the thing that generated the metric, would the metric continue? Organic word-of-mouth referrals continue without you. Viral growth from a genuine habit-forming product continues without you. Press-driven sign-ups stop the moment the story is off the front page.
Other common noise sources: NPS from early beta users who self-selected to believe in you, feature requests from a customer who is not paying enough to justify the request, demo-to-trial conversion rates that don't reflect whether trials become paying customers, and investor enthusiasm, which measures the market's appetite for the story you're telling, not whether the product does something irreplaceable.
What makes something a real PMF signal
A real PMF signal has three properties. First, it persists without your intervention. Second, it comes from people who have something to lose if they don't use the product. Third, it is behavioural, not attitudinal — measured by what people do, not what they say.
Retention is the clearest signal in any business. If the cohort survival curve flattens above zero — if users keep coming back six weeks, ten weeks, three months after first use — the product is doing real work. If the curve decays to zero, the product is not. Revenue retention tells the same story at the commercial layer: if customers expand, renew, and resist cancellation without aggressive intervention from your customer success team, you have signal.
Organic pull is the second signal type. This is inbound that you did not generate: a customer found you because another customer mentioned you, a founder reached out because they read a case study from someone in their network, a referral came in without a referral program. When organic pull starts outpacing your outbound effort, the market is beginning to pull the product rather than the other way around.
A triage framework for the metrics you already track
Apply this test to every metric on your dashboard. Write down the metric. Answer: does this metric require my active effort to sustain, or does it persist without me? If it requires active effort, label it a growth metric (useful, but not a PMF signal). If it persists, label it a PMF signal.
Then ask: is this measuring behaviour or attitude? Attitude metrics (NPS, satisfaction surveys, interview enthusiasm) have their place but should never be treated as PMF evidence. Behaviour metrics (retention cohorts, usage frequency, unaided return rate, churn, expansion revenue) are the ones that can actually tell you whether the market needs what you built.
Most founders, when they run this triage, find that their dashboard is 70-80% noise. That is not a problem to be fixed immediately. It is a diagnosis. The question is not how to improve the noise metrics — it is whether any of the remaining 20-30% are showing the right shape.
Score your own PMF in 20 minutes.
Free PMF score across market, founder, and execution readiness — with named blind spots and specific first actions. No credit card required.