The first metric I owned at Angel One was web DAU as a share of total DAU. It sat at 4%, and my job was to move it. Before I could, I had to answer a question nobody had asked out loud: what counts as a daily active user on web?
A visitor who lands on a public market page and leaves? A logged-in user who checks a portfolio and never places an order? Someone who opens the site inside the app's webview? Each answer produced a different number, and each number pointed at a different roadmap. Optimising for the first would have sent me chasing SEO traffic. Optimising for the third would have meant measuring the app and calling it web.
We settled on a definition, wrote it down, and the number moved from 4% to 7% over the year — but what moved it was the clarity, not the cleverness. Once everyone agreed what the denominator contained, it became obvious which users were missing and which surfaces they were missing from.
I have since seen the same trap in every dashboard I have been handed. Someone defined the numerator eighteen months ago for a reason that no longer exists, and a whole team optimises against that fossil. Nobody is lying. Nobody checked.
So the first question on any metric I inherit is not “is it going up?” It is “what exactly is being counted, and what exactly is it being divided by?” If I can't answer both in a sentence, I don't own that KPI yet — I'm just watching it.