Two weeks, four channels, zero signal. The problem wasn't the product — people couldn't tell what it did in 5 seconds.
Posting the numbers instead of a win, because the wins I read here didn't help me nearly as much as someone else's failure would have.
I built a tool that tells you whether an AI launch is real or mostly hype. You paste in a launch, you get one verdict, and every claim in it names its source. I shipped it and pushed it across four channels over two weeks.
Result: zero. Not "low conversion" — zero signups, zero emails, zero replies.
My read at the time: nobody wants this.
What was actually happening, and it took two strangers saying it before I believed them: people opened the site and could not work out what it did. They weren't rejecting the idea. They never got far enough to have an opinion about it.
Three things I got wrong:
-
Posts from an account with no audience reach nobody. My own numbers: a standalone post got 1-2 views. Comments left inside other people's active threads got 12-23 views each, plus real replies from real people. Identical effort, ten times the reach. I was posting into a void and calling it a demand test.
-
Silence is not data. I read "no signups" as "no demand". But if nobody understood the page, that silence was measuring my copy, not my market. You can't learn anything about an idea from people who never understood the idea.
-
One confused person is worth ten polite ones. "Looks cool" taught me nothing. "I don't get what this is" was the single most useful thing anyone said to me, and my first instinct was to argue with it.
What I did about it: built a 22-second explainer that plays right on the page. Four scenes, no video file, plain HTML/CSS/JS, so it can't 404 and needs no hosting. The first thing a visitor sees now is what the thing does, not what it's called.
Then I found out the explainer itself opened half off the bottom of the screen on a 1366x768 laptop, and started playing before you could scroll to it, so a visitor arrived at step 3 having missed the two scenes that explain what it even is. Fixed that too, and measured it across 14 screen sizes instead of eyeballing it, which I should have done the first time.
Whether any of it works, I honestly don't know yet. It's day one.
The tool: https://bharatlearner18-del.github.io/reality-filter/
The explainer: https://bharatlearner18-del.github.io/reality-filter/about.html
What I'd actually like to know from people here: how did you find out your landing page wasn't landing? Did someone tell you straight, or did you work it out from the numbers? I clearly can't see it myself, and waiting for a stranger to be blunt with me isn't a strategy.
The three lessons are all correct, and lesson 1 (comments in active threads beat standalone posts 10x) is the most useful thing here, it's exactly why this post got engagement and your standalone ones didn't. You already found the channel that works; you're using it right now.
On your actual question, finding out your page isn't landing without waiting for a blunt stranger: the numbers never tell you, and that's the trap you half-named. Zero signups is undifferentiated, same reading whether the idea is dead, the copy is confusing, or the page broke on a 1366x768 laptop. Silence is one output for a dozen failures, so you can't reverse it into a cause. You have to watch the confusion happen, not infer it.
The fastest method, and it's uncomfortable, which is why people avoid it: the five-second test. Show the page for five seconds, take it away, ask "what does it do and who's it for." Not "do you like it", just "what is it." Run it on a friend, a Discord, five people on a usability-test site for a few dollars. If they can't answer, you've found the leak in minutes instead of two weeks, before spending any outreach on a page that can't convert yet.
The deeper reframe: you ran a demand test on a page that hadn't passed the comprehension test. Two different experiments, and they happen in order. You can't measure whether people want the thing until they understand the thing. Every outreach effort before the page is comprehensible measures your copy, not your market, you said this yourself, it's the sharpest line in your post.
So the sequence: comprehension test first (5-second test, cheap, before any launch), then demand test (do people who get it want it), then conversion. You jumped to step 3's channel with a step-1 problem. Fix comprehension, re-run the exact outreach that got zero, and you'll get a clean read on demand.
To directly answer: I've never once worked it out from numbers. Always a person, watching them get confused in real time. The numbers tell you something's wrong; only a human tells you what.
"Two different experiments, and they happen in order" is the line I needed. I'd been treating comprehension as a copy problem inside the demand test, not as a gate in front of it.
Here's what happened after your comment, since a reply that doesn't report back is worthless.
I rewrote the headline. It used to be a question — "Is that AI launch real, or just fog?" — with the sentence that actually explains the product sitting underneath it in smaller type. It now reads "Type any AI launch. Find out if it's real or just hype." Action first, payoff second, product name nowhere.
I also cut a word I'd been in love with. I'd been saying "fog" everywhere, and it turns out the product's own interior already says "hype" — the risk meter inside a report is literally labelled Hype risk. So the headline was the only surface speaking a private language, and I was the last person who could see it.
What I haven't done is your actual advice. I haven't run the five-second test. I rewrote the page and I'm now sitting here believing it's clearer, which is precisely the position I was in last time. Doing that this week — a friend, a Discord, and a few paid strangers so it isn't only people being nice to me.
The line I'm stealing: silence is one output for a dozen failures, so you can't reverse it into a cause. That's my whole two weeks in one sentence.
The "I don't get what this is" line hits home. We had the exact same thing with a browser extension — two months describing it as "AI-powered engagement" and getting nothing back. The fix that worked for us was identical to yours: describe the moment of value, not the category. Input → verdict, in one sentence, before any name.
And your point about comments inside other people's threads getting 10x the reach of standalone posts matches our numbers almost exactly. Replies distribute like posts. Standalone posts vanish.
Ran your formula against my own page and it failed its own test, so: thank you.
Before — headline: "Is that AI launch real — or just fog?" Underneath, in smaller type: "Check any AI launch. Get a straight verdict, backed by evidence."
The second line was already doing input → verdict. The big line was just asking a question. I had the right sentence on the page the whole time and had ranked it second.
After — headline: "Type any AI launch. Find out if it's real or just hype." The evidence claim moved down to the support line where it belongs.
The other thing your comment cost me: "fog." Two months of my own vocabulary. The report itself already labels its meter Hype risk, so the headline was the only surface speaking a language visitors hadn't learned yet.
Two months on "AI-powered engagement" is a painfully familiar number. Did the rewrite move anything measurable for you, or did it mostly change who bothered to reply?