I spent two weeks optimising the wrong variable — how long did yours take to show up?

I audited my own video numbers this week and found something I'd rather not have found.

For two weeks I'd been rewriting titles. Moving keywords around, comparing which phrasing pulled better. I had a whole theory about it.

Then I opened the per-video tab and sorted by length. Short clips: 21 uploads, median 32 views. Long ones: 12 uploads, median 1. Same channel, same weeks. One title appears twice at two different lengths — 252 views at ten seconds, 5 views at two minutes. Identical words.

So the variable I could actually see was runtime, and the variable I was working on was grammar. About a third of everything I published went into the segment that returns nothing.

The uncomfortable part isn't the mistake. It's that I only found it because I happened to open a tab I'd never opened. There was no reason I'd have looked — the summary numbers looked fine, they just averaged the two groups together.

The obvious move is to stop making the long ones. I keep not doing it, because they're the only thing I actually enjoy making, and I can't tell whether that's a real reason or a preference wearing a suit.

How long were you optimising the wrong variable before something forced you to look? And was it a tool that showed you, or a person?