Problem

A product's life, from first idea to the customer, scattered across teams and eight tools that never fit.

Built

Our own product lifecycle tracker. Prototype in 30 minutes, done in four hours.

Status

running, quiet

Every product we make has a life. Someone has an idea. Design picks it up, R&D argues with design, a prototype gets built, someone works out what it actually costs, photos get shot, and somewhere along the way the whole thing gets revised two or three times, because that is usually its fate. Then it goes on sale and the story ends.

For years we couldn’t see that life in one place.

We tried. Excel, Airtable, ClickUp, and a handful of others whose names I’ve honestly forgotten. Eight serious attempts over the years, each with the setup time, the training sessions, the “this time it’ll stick” speech. Each one held a piece of the process well. None held all of it. Because what we needed was oddly specific: every team sees the whole flow, each person can only touch their own part of it, and the company keeps the history forever. Which idea became which product, who changed what, when, and why. No off-the-shelf tool gave us all three without a fight.

Then another meeting, and the topic came around again. It always comes around again; we put out dozens of products a year, so the wound reopens on schedule. Except this time I’d spent a while building tools for myself with AI, and I sat through the meeting thinking: maybe this one too.

Here’s what was different about this project. I’ve been inside this process for about twenty years. I don’t do every step myself, but I’ve run it long enough to know exactly how it flows and exactly where it stalls.

Still, I didn’t trust my own picture of it. Right after the meeting I pulled in everyone who touches the process and we drew the real flow on a board. Then we did an autopsy on the eight dead tools: where each one broke, what we worked around, what we quietly stopped filling in. And we talked about permissions, which turned out to be the actual spec. The flowchart was the easy part. Who is allowed to touch which box, that was the product.

I took all of it to ChatGPT and argued with it until the notes became a spec. Its questions were good; the spec came out sharper than our whiteboard version. Then I opened a terminal, pointed Claude at the spec, and half an hour later I had a working prototype. I pushed a few test products through it to see if the flow held. It held. Four hours in, the thing was done. Twenty years of watching that process, four hours to model it. Fastest build of my life, and it isn’t close.

Next morning I demoed it and handed it to the team. Week one went the way week one always goes: could it also do this, this bit doesn’t work, it would be nice if. The requests ran for about a week, then thinned out.

That was four months ago. Today the idea, the design, the R&D back-and-forth, the production prototype, the costing, the specs, the final photos, and the two or three inevitable revisions all live in one place, with history. The last feature request came in two months ago. In my ledger, silence is the best review a tool can get.

It stays in-house, and not because it’s precious. I don’t think it would sell. A tracker shaped this exactly is a cast of how one company works. Your company works differently, and that’s precisely why our eight store-bought attempts never fit us. I’d rather tell you to build your own than sell you a ninth tool that almost fits.

What I’m keeping from this one:

• Some processes never show up as a cost line. They show up as friction, arguments, and everyone being a little more tired than they need to be. Fixing one doesn’t move the P&L. The people notice anyway.
• The better I know a process first-hand, the faster I can model it. The AI did the typing; twenty years of know-how set the speed.
• Ask about permissions before features. Drawing the flowchart took an hour. Deciding who may touch which box was the real design work.

Built for myself first. This one, for everyone down the hall too.