What shipping internal tools taught me about analytics

Most of what I know about analytics, I didn't learn from a course. I learned it from watching people ignore things I built.

The first dashboard I made at work was, I thought, pretty good. Clean visuals, sensible filters, all the numbers in one place. Almost nobody opened it. When I asked around, the answer was simple: it answered questions nobody was asking. The reports people actually used were ugly Excel files — because those files answered the exact question someone had on a Tuesday afternoon before a review meeting.

That changed how I work. Now, before building anything, I try to sit with the person who'll use it and watch what they currently do — which spreadsheet they open, which column they check first, what they copy into WhatsApp. The tool almost designs itself after that.

The second thing: adoption is the real deliverable. A compliance platform that tracks five hundred obligations is worth nothing if departments keep their own spreadsheets on the side. Getting people to actually switch took more effort than building the system — training, follow-ups, small fixes that made their specific workflow one click shorter. Nobody claps for that work, but it's the difference between a project and a screenshot.

And third: AI-assisted development changed my maths on what's worth building. Tools I would earlier have put in the “someday, if IT has bandwidth” bucket can now be prototyped in days. The bottleneck has moved from writing code to knowing exactly what the business needs — which, honestly, was always the harder part.

I'll write more about specific builds soon. For now, the case studies under Work go deeper into each of these lessons.