Last week I wrote about never having written a line of code before June. This is the next instalment, and it's about the part nobody warned me about: getting one real person to use the thing.
Short version: no paying customers yet. But this weekend I learned more about why than in the four months before it.
1. We found a hook. Maybe not the right one.
My first ads were claymation films.
Then we tried a short video in the style of those castle-defence phone games, the Clash of Clans look: little soldiers defending a castle, with Enki as the memory keeping the walls up. It went past 200 plays in about a day. That told us we'd found a hook. Whether it's the right hook, one that brings in people who want what Enki does rather than people who like castle games, I don't know yet.
A follow-up, "Raid on Memory Castle", got 203 views from 166 people on Facebook. 58% were over 55 and a third were over 65. Average watch time: 7 seconds. Link clicks: 5.
I've put £30 behind it as a paid test (£3 a day for ten days), with a stop rule written down before it started: more than £1 a click and no sign-ups after four days, it stops.
2. The front door was locked
I finally pulled the server logs and counted. About 18 people reached the sign-up screen. None got past it.
The first screen asked for date of birth, email and an 18+ tick, all at once, before you'd seen anything. Safe, responsible, and a wall.
On Sunday I split it:
- Tick that you're 18 or over, give an email, you're in.
- Your birthday is asked later, inside the app, the first time you reach an adults-only part.
- If the birthday is under 18, it stops kindly and points to proper help.
While I was in there I checked whether the sign-in email could be hammered by a bot. It could. It's now five emails an hour per address, and when you hit the limit it tells you so instead of pretending an email is on its way.
3. I couldn't see any of it, so I built a dashboard
I was posting to Facebook groups, Instagram, LinkedIn, Bluesky and Indie Hackers and losing track of what went where. So I built a small desk that:
- logs every post
- shows which links actually got clicked
- nudges me when a scheduled post is due
- flags when the overnight tests go red
It's how I found the 0 out of 18. Without it I'd still be blaming the ads.
What I'd love your help with
- If you've found your first ten users, where did they actually come from?
- Has anyone found their audience was older than they assumed? What changed when you leaned into it?
- Is £30 on a video test a sensible first experiment, or noise at that size?
- What would make you trust an AI companion with your memories?
I'll answer every comment. If you want to see it: https://enki-companion.onrender.com/?utm_source=devto&utm_campaign=ep1_first_customer
Thanks for reading, from a 56-year-old who is now more scared of a sign-up form than of code.
Top comments (3)
Splitting that first screen was the right call. Asking for date of birth before someone has seen anything is a lot, and the 18+ tick now with the birthday asked later is a fair way to handle it.
One thing on the email limit: five an hour per address stops someone hammering one inbox, but a bot can just rotate addresses. Worth adding a cap per IP as well, so the same source can't fire sign-in emails at hundreds of different inboxes, run up your email bill and get your sending domain flagged.
Also good that it tells people when they've hit the limit instead of pretending an email is on its way. That kind of thing quietly loses real users.
Thanks Elijah, that's a really useful catch. You're right: five an hour per address stops one inbox being hammered, but it does nothing about one source rotating through lots of addresses. There's a general per-IP rate limit on the whole app, but nothing specific to sign-in emails, so a per-source cap on those is going on the list now, along with keeping an eye on the sending domain's reputation.
And agreed on the honest "you've hit the limit" message. A fake "it's on its way" is the kind of thing nobody ever reports, they just leave.
Your dashboard did the most important work here: it stopped you blaming the ads and showed the front door was the problem. Splitting the form was a good fix, because asking for a birthday before someone has seen any value is a high price for a first step. One test I would add is a stranger test. Open the form in a fresh browser with no saved details, try to sign up, and write down each step: reached the form, submitted it, saw an error, got the email, got inside. If any step has no log line, add one, because a funnel with a missing step is how a 0 out of 18 stays invisible.