Signups feel like progress. They are not the number that tells you whether the product will sell.
Plenty of AI products get signups and no payers. One founder reported 500 signups in 30 days and 3 paying customers. Another reported 70 signups and none. The gap has two possible causes, and each one needs a different fix. This test tells you which one you have.
Signups are the wrong number to watch
A signup is a curious person, not a buyer
A free signup means someone was interested enough to click. It doesn’t mean they have the problem badly enough to pay. A free tier that works well can make this worse, because the person gets enough value to stop there. One founder who put a free scan in front of his product found that people saw the gap and the fix at the same time, then went off to fix it themselves.
Count conversations, not clicks
For this test, ignore signups. Track two things each week: real conversations with people who used the product, and paying customers.
Run the one-week test
Count real conversations
Aim for ten conversations in one week. A real conversation is a call, a reply that answers your questions, or a half hour of messages. A click is not a conversation.
Ten is a useful target, not a scientific cutoff. Fewer than five in a week means the problem is reaching people.
Count the “interesting, but” replies
Then read every reply you got. Mark each one that says “interesting, but” followed by a reason they wouldn’t pay. “Interesting, but I’d need it to connect to my CRM” is one. “Interesting, but I’m not sure it’s worth the money” is another.
Most conversations ending this way means the message is the problem.
What each result tells you
Few conversations: a distribution problem
If you had fewer than five conversations, the people with the problem either don’t know you exist or don’t see the link between your product and their problem. The fix is reaching them where they already complain about it. Changing the product won’t help yet.
Many conversations, no payment: a message problem
People who talk to you but don’t buy understand the product and still don’t see why it is worth paying for. The fix is the sentence that says what changes for them. A new feature won’t fix it.
Both: start with distribution
When both numbers are low, fix distribution first. A better message reaches nobody.
What this looks like in practice
Maya is a composite, built from patterns in the founder posts above. Her numbers are illustrative, not data from one company.
Week one: the numbers
Call her Maya. She built a tool that drafts client status reports for small agencies. In her first month she had 140 free signups and no payers. She had four conversations, and three of them were with people who had signed up on their own.
Four conversations is under five, so the test points to distribution. Her signups came from one launch post, and the people she needed weren’t reading it.
What she changed
She stopped the launch post. She asked the four people she had talked to where they go when a client report is late. Two named a paid community for agency owners. She posted there once a week, answering questions about client reporting before she mentioned the tool.
Three weeks later she had eleven conversations. Seven were “interesting, but” replies. Most said the tool was useful but they didn’t see what it would save them. Better distribution had exposed the second problem.
The fix to try first
One change per week, measured
Pick the single fix that matches your result. Make one change, run it for a week, and count again. Change two things at once and you won’t know which one worked.
Maya’s next change was the product page. She led with the hours saved instead of the feature list, then counted “interesting, but” replies again the following week.
Next step
Read the framework first
If you haven’t read why products stall in the first place, start with Your AI product has no users. The problem is not the product. It names the three reasons. This test tells you which one applies to you.
Run the test before you spend another month on either fix. If the numbers are still unclear after two weeks, a positioning review is worth paying for. We do that work, and it starts with the same count.
