118 sign-ups, zero attendees. Week of 24–30 August
· 8 min read
On Thursday, 27 August, I ran a free webinar. It had 118 registrations. Nobody came.
Not "few." Not "half." Zero — all 118 rows in the export say no-show.
There were five people in the room. None of them are on that list: they got there through my DMs and Telegram, bypassing the official funnel entirely.
This is the first piece on this blog, and I'm not opening with a shop window. I'm opening with the most expensive thing I own from August — an honest number. What follows: what it actually means, what I changed because of it, and the three rules I took out of the week.
A flat zero is not "they didn't want to come"
A free webinar normally converts in the tens of percent. A bad one converts in the single digits. A clean zero across the entire platform, while five people walked in through DMs, reads differently: they didn't decline. They never got a working path into the room.
Here's what I found when I dug in:
- Registration lived on the platform. The room link and the reminders lived in my DMs. Those two loops never once connected.
- Registrations had been piling up for three months: 52 in June, 21 in July, 45 in August. The last one arrived on the day of the event. The page had been collecting people all summer — it was never opened for a specific date.
- The event was titled "AI Marketing specialist." That title matches neither the topic of the 27 August session nor the course behind it.
Someone signed up in June for "AI Marketing specialist," and at the end of August nobody invited them to a session about building an AI employee. They didn't miss it. They weren't asked.
This gets fixed in settings, not in content. Which is honestly good news: a broken pipe is cheaper than a broken product.
The number that hurts more
For the first time in three months I have the make-up of my audience instead of a feeling about it.
| Who | People | Share |
|---|---|---|
| Students, school kids, interns | 30 | 25% |
| Employed IT/AI specialists | 28 | 24% |
| Company employees | 17 | 14% |
| Startup team members | 17 | 14% |
| Tech community | 11 | 9% |
| Investors, teachers, other | 15 | 13% |
And the key fact: 82% of registrations came from personal email. 83 Gmail addresses, 14 Mail.ru. Corporate domains across the whole list: about eight.
These are individuals paying out of pocket, not companies with a training budget. A quarter of the list doesn't pay for anything, ever. Another quarter are salaried specialists for whom my course price equals one to three months of their personal learning budget.
The real core of my audience here is those 17 startup people. Fourteen percent. Not 118.
For three months I thought I was building a launch list. I was building a list of the wrong people.
The caveat, without which that conclusion is a lie
The temptation is to write "they can't afford it" and close the case. You can't, and here's why.
Zero out of 118, at 95% confidence, does not mean "conversion is zero." It means "conversion is no higher than about 2.5%." And if you count honestly — only across the 17 who actually resemble my customer — then zero out of seventeen, off a single session, is very weak evidence. Close to nothing.
So the decision "stop addressing the course to this list" is correct. The sentence "this audience doesn't pay" is not a fact. It's a guess I haven't tested yet.
The gap between those two sentences is the gap between working with data and bending data to fit your mood. I catch myself doing the second one far more often than I catch myself in time.
Decision two: LinkedIn is out of launch warm-ups
2,887 followers. The week: 778 impressions, 354 people, 5 reactions, minus one follower.
Russian-language posts, week over week: 215 → 153 → 97 impressions. Inside the last week, post by post: 101 → 66 → 50 → 25. (Two adjacent exports, which is why the first post of the week shows up in both.)
And the best post of the week wasn't a post from that week at all. It was an English one published the week before, pulling 160 impressions on its tail. Nothing I wrote for the webinar came close.
The demographics explain all of it. Among my followers, Moscow is 35%. Among the people the algorithm actually shows my posts to, Moscow is 2% — while the Bay Area is 25% and New York 15%. I'm writing in Russian for an audience the platform doesn't serve me.
The decision, in writing: LinkedIn no longer takes part in launch warm-ups. Not for webinars, not for the course. One Russian post a week stays — a personal story, for reputation, with no expectations attached and no more than five minutes spent. The English feed continues as it is: it earns reach for free and gets in nobody's way.
One more thing, a format I nearly ruined by transplanting it. On Saturday I filed "the backstage post" — an honest write-up of how something actually works, screw-ups included — as a winner, and proposed making it a weekly column. On Sunday the numbers arrived and I had to withdraw that: the post-webinar backstage piece did 50 impressions and zero reactions on LinkedIn. The same move on Telegram: 43 views against 68 subscribers — 63% of the channel, best post of the month.
A format doesn't "work" or "not work." It works on one platform and does not transfer to another. I managed to write down the wrong conclusion, and I managed to take it back — because I waited for numbers instead of staying with a feeling.
What broke quietly
One more thing from the week that's easy to miss: I stopped being able to see Telegram.
All these months I'd been reading view counts off the public web preview of my channels. This week it rolled back to a late-July cache and has been showing stale numbers since. The workarounds closed one after another.
Which means I never had a measurement channel of my own. I was living on a free thing nobody ever promised me, and it broke in exactly the week I most needed to measure the return.
That's probably the most portable lesson here: if you don't control the source of a number, you don't have a number. You have a habit.
And the plain misses
Four of seven planned posts went out. The "starting at 6pm, last call" post never fired on the day of the event. The Saturday post wasn't in the queue at all. I found out about neither when they failed — I found out on Sunday, checking the plan against what had actually shipped.
The rule I wrote from that: on the day of an event, the post goes out by hand and gets confirmed with my own eyes. Critical posts don't get entrusted to a scheduled queue nobody has checked in a week.
What I changed for this week
I've closed the free webinar as a format. Not because it's a bad format, but because a free registration measures nothing: it costs zero, so it weighs zero. A hundred and eighteen of them is a beautiful number with not one fact behind it.
In its place: a paid workshop. Small ticket, ten seats, ninety minutes. Payment by hand, in DMs, with no checkout page — by hand first, until you understand the shape of the thing.
And I wrote the success criteria before the test, not after: how many payments mean "the price is too low," how many mean "the core is real but narrow," and under what outcome the conclusion isn't about price at all but about the email failing. A criterion written in advance is the only defence I know against declaring any result a win.
The short version of the week: stop counting registrations, start counting people who paid for something.
Three rules I'm taking with me
- A registration is not a person. Count the ones who showed up. Free intent is worth nothing precisely because it cost nothing.
- A flat zero across a funnel is a hypothesis about plumbing first, and about people second. It's almost always operations, not the market.
- A number whose source you don't control cannot sit under a decision.
What's next
I write one of these every week, on Monday — the same numbers, whatever they turn out to be. The next one covers the paid workshop, including how many of those ten seats actually sold.
If you're running the same experiment — a solo founder trying to tell a real signal from a flattering one — I'd like to hear how it's going for you. Message me on LinkedIn; I read everything.