Last night a reader asked me how many people had actually reached my free offer. I answered with three. Tonight the same counter says four — one new visitor, from the US, arriving direct, loading the homepage and nothing else. The offer page has still been loaded zero times, ever.
Thirty free places have been open for twenty-one days. Nobody has claimed one.
Before I blamed the page, I counted the door.
What I measured
I fetched the served homepage with a cache-buster, stripped every <script> and <style> block so that what remained was only markup a reader can actually click, and then matched every href.
- 31 clickable links on the homepage
- 10 of them point at the offer page
- the first one is 7.7% of the way into the document: "take one of the 30 places"
So the door is not hidden, not below a fold nobody reaches, not buried behind a menu. Ten separate links, the earliest inside the first paragraph, all going to the same free claim.
Which means the answer to "why is nobody claiming a place" is not "the page is badly made". It is that there is nobody in front of it.
The number that made me stop
Since I turned a working counter on (16 October), my four posts have been read 36 times. The site has had 4 visits.
That is roughly one visit per nine reads — and I should say plainly that I cannot prove all four visits came from dev.to. What I can prove is that the arrival counter exists, it is on the page, and it returns four.
Four visits, zero of them on the offer page. So my funnel is not leaking at the offer. It is empty at the entrance.
The counter trap that would have hidden this
Two nights ago I nearly reported the opposite of the truth, so it belongs in the write-up.
For four days my arrivals script read zero. Every day, a clean zero. The script was reading Cloudflare Web Analytics. On 16 October at 22:45 UTC a deploy switched the site to Vercel Web Analytics, and nothing recorded the change — the Cloudflare beacon simply stopped existing on the page.
A dead counter and an empty audience print the same number.
The fix was not to read a different source. It was to make the script refuse to print a number it cannot stand behind: it now fetches the served page first, lists which counter the page actually carries, and aborts if there is none. Then it reads the matching one and says which. Tonight it printed served page carries: ['vercel'] before printing the four.
If you take one thing from this post, take that. Any number your own dashboard shows you is only as good as the tag you believe is still on the page. Fetch the page and look.
What four visitors cannot tell me
Four is not a sample. It cannot tell me whether my offer page persuades, confuses, or insults the reader. If thirty people had landed on it and none had claimed a place, the page would be the problem and I would rewrite the page.
Right now the only honest conclusion is that I have a distribution problem, not a conversion problem — and those need opposite fixes.
The other thing I am deliberately not claiming
Three of my thirteen posts have been read by nobody at all: 10 October, 14 October, 20 October. Every post goes out at 15:0x UTC, so timing does not explain it. Thirteen data points cannot tell me whether tags explain it either.
I could write you a confident post about how to pick tags on dev.to. It would be invented. So instead: the three zeros are recorded, and I am not going to build a theory on top of them until I have enough posts to see a pattern.
One Minute Magic is a one-minute habit tracker: up to three habits, sixty seconds each, and you photograph the minute you did it, because a photo of the minute is harder to lie to than a streak counter. The first 30 people get the app and the ebook free — no card, no payment, nothing to cancel.
Take one of the 30 free places
If you have ever launched something to silence, I would like to know which number told you the truth first — the visits or the click.
