The Free Trial Is Lying To You
Everyone obsesses over trial conversion. I think the number is mostly noise — and chasing it got me building the wrong things for two years.
Every bootstrapper I talk to can quote their trial-to-paid conversion rate within half a percent. Mine's 11. Mine's 14. Mine's 8 and I'm panicking.
I did this too. For about two years I treated that number like it was the heartbeat of the business.
Then I actually looked at who was signing up for the trial, and the number stopped meaning anything.
Happierleads sits in a category where half the trial signups are competitors doing recon, students writing a thesis on identity graphs, and people who read a LinkedIn post and wanted to see the dashboard. None of them were ever going to pay. All of them count against the denominator.
So the "conversion rate" is really two numbers smashed together: how good I am at closing real buyers, and how much random traffic showed up that week. When I chased the combined number, I ended up building things to convince people who weren't buyers to become buyers. Which is the worst product work you can do.
The metric I watch now is conversion among people who hit the activation moment in the product. Installed the pixel, saw a real identified visitor, came back the next day. That cohort converts at a number I won't quote, but it's roughly 5x the headline rate, and it actually responds to things I ship.
If that number moves, I did something. If the headline number moves, Reddit linked to me.
What I'd do this week if I were you:
- Define one activation event in your product. One. The moment where a trial starts behaving like a customer.
- Measure conversion only inside that cohort. Ignore the rest for product decisions.
- Stop reporting headline trial conversion in your own head as if it's a KPI. It's a vanity ratio with a fever.
- Spend this week on the step before activation, not the email after signup.
The headline number is for investors you don't have. The cohort number is for you.
Talk soon,
— George
P.S. I spent an hour yesterday arguing with my own support AI about whether a bug was a bug or a feature. It won. I'm not sure how I feel about that.