Let me tell you a story about two teams working on the same mental health app.
The first team builds a mood chart. This where people record how they feel each day, then look back over the week to see if anything has changed and reflect on what caused the changes.
The chart shows a line and coloured points. But when QA tests it with a screen reader, all they hear is that there's a chart. They can't tell which days were difficult or read back what they recorded. So they send it back for improvements. The dev team spends a while on this and they decide to add a short written summary of the pattern and let people read each day's entry as text.
QA tests it, revises it and then it goes live.
A few months later, another team builds a sleep chart.
It looks different from the mood chart and it also has a different purpose when planning it, so nobody thinks to look at any of the earlier work. During testing, QA finds the same problems. Someone using a screen reader can't tell whether their sleep has changed over the month and what caused it.
The same film plays again. The team starts asking questions the team that built the mood chart had already answered. What should a summary include? How can someone check the individual nights? What should happen when the data changes?
By the time someone mentions the mood chart, the sleep team has spent days working through those questions. They could have used what the mood team learned. They would have still needed to make their own decisions about sleep, of course.
Instead, they rehashed the same discussions during meetings, Slack messages and Gemini transcripts. Of course, they still didn't record anything. You should see by now there's a problem here, because a few months later some other team will build another chart and work through the same problems.
Nobody is doing anything careless here. They're actually very efficient in finding problems and fixing them. The first team solved the problem in front of them. The second team couldn't use work they didn't know existed. And the next team will go through the same motions.
That’s the bit that bothers me. The company paid for the same learning twice and, unless something changes, it'll pay again.
What should change then? The simplest fix I can think of is for the decisions to be recorded as they are made. I'm not asking anyone to write a handbook. I just want the next person building a chart to find out that someone else has already wrestled with this. What did they try? What worked? What would they have done differently?
The sleep team still has to make its own decisions. But at least it can start where the mood team left off.