claimlead is dead. It ran for about a month. I am writing this down while the numbers are still in front of me, because post-mortems written later tend to get flattering.
What it was
A board where products paid to hold a rank. That already exists — outbid.lol is the one people know — so the difference was supposed to be honesty: every listing published its own click count and cost per click, live, including when the number was embarrassing.
It turned out the number was embarrassing.
The numbers
- 202 visitors, lifetime
- 7 clicks, lifetime
- 1 listing — and it was mine
- The launch was 5 tweets over two days. They got 0 likes, 0 retweets, and 0 replies from anyone other than me.
That last one is worth sitting with. Not "low engagement". Zero. A throwaway tweet I posted in January got two likes, so the account was not shadowbanned — those five posts simply reached nobody who cared.
The mistake was made before I wrote any code
I never checked whether anyone was searching for this.
Not "I checked and the numbers were marginal". I did not run the check at all. I had an idea I found interesting, I could see the shape of the product clearly, and building it was more fun than validating it. Everything downstream — the copy, the launch, the outreach plan — was work spent on a foundation I had never tested.
The comparison is uncomfortable, because I have since done it properly for the product I am building now: I found roughly 6,650 searches a month across a cluster of terms, at low difficulty, before the first commit. Same person, same month. The only difference is the order I did things in.
The structural problem: nobody on either side
A board needs listings to attract traffic and traffic to attract listings. I had neither, so I tried to bootstrap by offering free ranks to eight companies I admired — Plausible, Buttondown, Tally, and others.
Read that back as the recipient: a stranger emails you, offering you a free spot on a board you have never heard of, with 202 lifetime visitors. You are not being offered traffic. You are being asked to lend your name to an empty room so it looks less empty.
That is not an outreach problem you can fix with a better email. It is the product asking other people to solve its cold start.
The email was good. It did not matter.
This is the part I actually want to remember.
The mail was, I think, genuinely well made. It opened by naming the two objections the reader was already forming — you do not know me, and this looks like an outbid copycat — because naming them is the only move that buys attention. It published the embarrassing traffic numbers instead of hiding them. It asked for something small. It promised no follow-up.
None of that mattered, because the offer underneath it was worthless. Good copy cannot rescue a product with no demand. It can only make the failure more articulate.
I now think craft applied to an unvalidated premise is one of the more expensive ways to waste a month, precisely because it feels like real work.
What actually survived
More than I expected:
- The codebase. The whole thing was forked into what I am building now. Two weeks of frontend, auth, billing and deploy that I did not have to write again.
- The board itself, repurposed as the house-ad rotation on another site I run.
- The honesty mechanic — publishing your own numbers even when they are bad. It was the best idea in the product. It was attached to the wrong product.
Failed projects are not usually a total loss. They are usually a working component attached to a premise that did not hold.
What I am doing differently
- Demand evidence before code. Not a feeling that people want it. A number, written down, before the first commit.
- Never build the second side of a marketplace first. If the product only works once other people show up, I need a reason they would show up on day one, and "I will email them" is not a reason.
- Stop treating an announcement as distribution. Five posts to an audience that does not exist is not a launch. Building an audience and launching to it are two different jobs, and I keep trying to do the second without the first.
This journal is me starting on the first one.
If you are about to build something: go and check the search volume for it right now, before you open the editor. It takes ten minutes. I did not do it, and it cost me a month.