When a feature you buried outperforms, do you reorder the page or wait for more data?

I went through my own view counts this week and found something I don't know what to do with.

One video is taking 246 of my last 355 views. The next two took 31 and 20. The one that's winning is about the last step of the process — the functional, unglamorous part I treat as a footnote everywhere else. Everything I actually lead with is sitting in the two that got 51 between them.

It's one video and I know how thin that is. But close to five to one against the next two combined is either a signal or the algorithm sneezing, and I can't tell which from where I'm standing.

The two things I could do:

Reorder around it. Move the last step to the front of the page and rewrite the headline around it. I'd know inside a week. Cost: if it was noise, I've rebuilt my positioning on a fluke and thrown away the line that actually describes what the thing does.

Leave it and gather more. Cost: if it's real, I spend another month leading with the wrong thing, and everything I make in the meantime points at the wrong end of the product.

What I can't get past is that I can measure which video got watched and not why anyone opened it. A title can win because the subject is what people wanted, or because the title happens to contain a word that gets typed a lot. Those two look identical in the numbers and they point in opposite directions.

For anyone who's been through this — what's the smallest change you'd make first? Do you move the whole page, or test the claim somewhere cheap and leave the page alone until it answers?