A founder with thousands of free trial users asked a familiar question this week: how do I know which of these people are actually close to buying? The honest answer is that most of them are not, and the trial numbers that look best in a dashboard are the worst at telling you who is.
This article is about reading intent from behavior. Not because behavior is fashionable, but because it is the only thing trial users cannot fake out of politeness.
Why don't signups predict revenue?
A signup costs the user almost nothing. An email address, a click, thirty seconds. Behaviors that cost nothing carry almost no information, because people perform them for curiosity, boredom, or politeness as often as for need.
The same is true of logins and page views. They tell you someone was in the building. They do not tell you whether the visit meant anything. Corporate AI pilots fail at reported rates above 40 percent for exactly this reason: they were green-lit on excitement, a free behavior, and killed on usage, a costly one.
What behaviors actually signal buying intent?
The signals that predict payment share one property: each one costs the user something real.
Inviting a teammate spends reputation. Nobody pulls a colleague into a tool they expect to abandon. Importing real data spends effort and trust, because now your product holds something the user cares about. Returning without a reminder email spends attention, the scarcest resource a trial user has. Hitting a usage limit twice in one week spends patience, and signals the product has entered a real workflow.
Every product has its own version of these. The question worth asking is: what does the invited their boss moment look like in my product? That is usually the intent signal.
How is stated interest different from revealed intent?
Stated interest is what people tell you: survey answers, positive onboarding feedback, a yes to would you pay for this. Revealed intent is what people show you through actions that cost them something.
The two diverge constantly, and the gap is where products die. Users are polite. They tell founders what founders visibly hope to hear, then quietly never return. A trial funnel full of kind words and empty of costly actions is not early traction. It is applause on the way out.
How do you build a simple intent score?
Start embarrassingly small. List the three to five costly actions available in your product. Score each trial account by how many it has taken in its first week. Then look at your existing paying customers and check which of those actions they took while on trial. Keep the actions that separate payers from non-payers, drop the rest.
This takes an afternoon with a spreadsheet. It does not require a data team, and at trial volumes below a few thousand accounts a data team would be reading the same tea leaves anyway.
What should you do with high-intent users once you find them?
Two things, in order. First, talk to the accounts sitting just below the payment line: high intent score, no purchase. They know exactly what is missing, and unlike churned users they still care enough to tell you. Second, resist the urge to blast the low scorers with discounts. A discount converts polite people into refunds. It does not create intent, it only borrows against it.
The larger habit underneath all of this: trust what strangers do over what anyone says. It is the cheapest research program available to a founder, and the least practiced.
Key takeaways
- Signups, logins and session counts are free behaviors, and free behaviors carry almost no information about buying intent.
- Real intent signals cost the user something: inviting a teammate, importing real data, unprompted returns, hitting limits.
- Stated interest diverges from revealed intent because people are polite. Build on what they do, not what they say.
- A useful intent score is a spreadsheet afternoon: three to five costly actions, validated against what your payers did on trial.
- Talk to high-intent accounts that did not buy. They are the cheapest, most motivated research panel you will ever get.