The usual advice — track retention curves, watch a Sean Ellis survey score, wait for the 40% “very disappointed” threshold — assumes you already have enough users for any of those numbers to be statistically meaningful. With ten or twenty early users, a retention curve is noise. What's actually useful this early is qualitative: specific things people say and do that are hard to fake and easy to miss if you're not looking for them.
Signals worth tracking before you have metrics
- People describing your exact problem, unprompted, in public.Not answering a survey about your product — posting in a community about the pain point before they know your product exists. This is the strongest signal there is, because there's zero incentive to say it if it isn't true.
- “Does this exist yet?” threads.Someone describing the exact shape of your product and asking if anyone's built it — a direct, unsolicited statement that the market wants this to exist.
- Frustration with the current best alternative. Complaints about the closest existing tool or workaround, specifically about the part your product does differently — tells you the wedge is real, not imagined.
- Unprompted feature requests from actual users.Early users asking for more of something, rather than you having to convince them to use what's there, is a much stronger signal than praise.
- People using it in a way you didn't design for.A sign the product solved a real problem well enough that someone stretched it, rather than abandoning it the moment it didn't fit their exact use case.
Where these signals actually show up
The first three happen almost entirely outside your own product — in the same public communities where people describe problems before they've found a solution: Reddit, Hacker News, niche forums for your specific audience. See the monitoring guidefor how to actually watch these continuously rather than checking occasionally, which is how most of this signal gets missed — the threads that matter are one search away, but only if you're looking on the right day.
The last two only show up once you have users at all, which is its own case for finding the first ones through the channels above rather than waiting — see the first-users playbook. Getrivescores exactly this kind of unprompted, public problem-description across Reddit, Hacker News, and a few other communities, so the signal doesn't depend on remembering to go looking for it. Try it free.