I shipped a paid product on October 5th. Thirty days later: 3 indexed pages became 91, three months of search impressions went from 13 to 115, and the number of people who completed the form was zero.
This is what I learned reading the numbers carefully instead of optimistically.
The funnel, with real numbers
| Stage | Count |
|---|---|
| Real visits (30d) | ~123, of which ~33 were datacenter/bot |
| Started the form | 2 (both on day 3) |
| Completed the form | 0 |
| Paid | 0 |
Two things stand out. First, 123 visits in a month is nothing — that's not a funnel problem, that's a visibility problem. Second, both form starts happened on the same day, three days after launch, which is exactly when an initial wave of "people who saw the launch post" would arrive.
The mistake I made reading this
For the first two weeks I treated "0 conversions" as a conversion problem. I rewrote lock-screen copy, added a download button, added draft persistence, fixed a tax disclosure. All of that was real work and some of it was genuinely broken — but none of it could move a number that was gated on traffic I didn't have.
The lesson generalises: optimising conversion when N is small enough that Poisson noise dominates is not work, it's fidgeting. With 123 visits you cannot distinguish a 3% conversion rate from a 0% one.
What I did instead, and whether it worked
Content depth. I compared my worst-ranking page to the top five competitors on the same query:
| Site | Body text |
|---|---|
| competitor A | 21,746 |
| competitor B | 16,115 |
| competitor C | 10,971 |
| mine | 3,653 |
I was ranking ~82 with 17% of the median competitor's content. Expanded four pages to roughly double their length with real examples and FAQ blocks. Over the following days: indexed pages 62 → 91, impressions 50 → 115, and the specific keyword went from position 82.6 to 76.0.
That's a real movement and it's still page 8. Movement isn't the same as progress, but it's the first evidence the direction is right.
Backlinks. The site had two backlinks in total. Two. That's the actual reason nothing ranks — not the content, not the tech, not the copy. I spent a day testing ~25 directories and content platforms and found that:
- Every AI directory has moved to paid ($47–$300)
- Most content platforms give
rel="nofollow" - One developer platform gives dofollow and has a login flow that works
Ended the day at ten backlinks, seven of them dofollow. Compared to two, that's the largest structural change of the month — and it will take 2–6 weeks to show up, if it does.
The uncomfortable parts
1. I had a bug that only affected paying customers, for weeks.
After checkout, the user was redirected to a page that verified the payment and then sent them back to an empty form — because the generated drafts were only in useState, and the redirect threw them away. Any customer would have had to refill nine questions and wait another minute.
I found it by reading my own code, not by using the product. The reason nobody reported it is that nobody bought.
2. My "distribution" was a log file.
An automated posting script printed posted nightly for eight days. Five of those days had posts. The account it was posting from had been silently placed in read-only mode, and my script treated "composer exists" as "post succeeded." Even in the days it worked, those posts got roughly 3 impressions each.
The distribution channel wasn't underperforming. It wasn't operating.
3. The marketing plan was written before the data.
There was a plan with phases: launch, then content, then paid acquisition. Written when the assumption was "we'll get some traffic from the launch and learn from it." The launch produced three visits. Everything downstream of that assumption was building on nothing.
What I'd tell someone at day 30 with zero customers
Check whether your channels are real before optimising them. Not "is it running" but "did an external system acknowledge it." A post with no permalink didn't happen. A directory submission with no confirmation page didn't happen.
Look for bugs on the paid path specifically. Read the code from payment → unlock → delivery. Walk it as a customer. Nobody else is going to.
Compare against competitors on your worst-visible keyword. Not your best. Your best is ranking for a reason. Your worst-visible keyword is where the diagnosis is.
Count your backlinks. If it's under twenty and you're ranking on page 7–9 everywhere, stop looking at your own site. The problem is off-site.
Where this is going
There are two weeks left before a checkpoint I set for myself. The criteria: 30+ indexed pages (done, at 91), 5+ daily impressions on long-tail queries (currently ~1.3), at least one completed form (currently 0).
The honest read is that I'll miss two of three, and the reason won't be effort — it'll be that a thirty-day-old domain with ten backlinks doesn't rank, and no amount of polish changes that.
Which raises the question worth being explicit about: is the lesson "I executed badly" or "the timeline was wrong"?
I think it's mostly the latter, but I've been wrong about that before. If you've been through this, I'd genuinely like to know which one it was for you.
(The product is Speech Kobo. The one thing I built that's unambiguously useful regardless of how this goes: an open dataset.)
Top comments (0)