Sample review runs on GMV. Sales don't.

The question

Every day, operators decide which creators get a free sample. The call is quick and runs on experience: open the request, glance at followers and GMV, approve or reject. A busy team reviews thousands a week.

But once a sample is approved, almost nobody goes back to check whether it was the right call. Review, shipping, content and orders live in different places and are hard to connect. So we connected them, and followed every approved sample all the way to whether it sold.

Review runs on GMV

Operators' calls follow a clear pattern. We trained a model on past review decisions and had it judge new requests: it matched what operators actually decided about 85% of the time. And what drives those decisions is clear too: about two-thirds comes down to which store it is and how much GMV the creator has.

Most samples don't sell right away

Within two weeks, fewer than 6% of approved samples led to a sale of the product requested. And sales are very concentrated: less than 1% of requests produced two-thirds of the units.

In other words, the return on samples comes from a handful of creators who really take off. The whole game is spotting them in advance, and GMV can't do it. Rank approved creators by how strongly review favored them, and the top sells at about the same rate as the bottom. No better than picking at random.

The best signal is in your own store

What actually predicts a sale is simple: whether the creator has already sold for your store. Creators who sold 20+ units with you in the month before asking sold again more than a third of the time. Those who had sold nothing: under 5%.

Operators usually approve these creators anyway. The problem is that this number isn't on the review screen. The card only shows GMV added up across every store, so operators are left using GMV to guess whether a creator will sell for them.

We checked

You might ask: creators who already sell for you probably have higher GMV too, so isn't this just GMV again? To rule that out, we compared only creators in the same store with GMV in the same range. Those who had sold for the store were still about 5× more likely to sell. Small creators or big ones, the result held.

And two weeks is only an early read. Orders keep coming in after that, so the real gap only grows.

How OperaX runs sample review

Experience only gets sharper with feedback. Sample review has never had any.

Operators aren't the problem. They review fast and consistently. What they've never had is a way to see what happened after they hit approve. That's what OperaX adds:

• See the result of every sample. Review, shipping, content and sales in one view, so it's obvious which calls paid off and which didn't.
• Put the right number on the review screen. Sales in your own store can sit right next to GMV, and proven sellers can be approved first.
• Recommendations that learn from results, not just from past review habits. That's what we're building into review now. AI suggests; your team decides.