We read the price of 72,898 AliExpress listings every day for a month. 89.8% of them moved.

I build a tool that records AliExpress prices once a day and stores each reading with its date. A month in, one number keeps surprising the people I show it to.

Of the 72,898 listings we track, 65,462 changed price at least once in 34 days - 89.8%. Only 7,436 held a single price the whole time. There was no sale event in that window.

That breaks the mental model most shoppers have, which is that prices sit still and drop on announced sale days. For most of the catalogue the price is simply in motion all the time.

Two things I am genuinely unsure about, and would like other makers' read on.

1. How do you present a finding that undercuts your own category? Half the tools in this space exist to tell people when the sale is coming. Our data suggests waiting for an announced sale date is often worse than buying on the Tuesday you first looked. That is useful to a shopper and awkward for the industry.

2. When is a short record honest to publish? We have 34 days. We cannot say anything about November's sale because we did not exist last November. I have been publishing the limitation as prominently as the finding, on the theory that "we do not know yet, and here is what we are recording so we can answer it later" builds more trust than a confident guess. Is that naive, or is it the only version worth publishing?

Happy to share the method or the shape of the data if it is useful to anyone.