October 26. I pulled a batch of listings that had just tripped our below-market flag — the badge that goes up when a car's asking price sits well under what its region cohort says it should. Out of the roughly 12,800 we flag across the system on an average day, I picked about 40 to actually watch: a mix from the UAE, a couple from Poland, a handful from Canada. Nothing scientific about the sample. I just wanted to see what happens to a deal after the badge appears, instead of writing about the moment it appears, which is the part everyone already covers.
My working theory, going in, was boring: good deals get bought fast, so the badge is basically a countdown clock. I was only half right.
October 28. Two days in, three of the forty were gone already — delisted, no trace. A 2019 sedan out of Dubai that had been priced low enough to flag went first, which tracked. Cheap, clean, well-photographed. Not a surprise.
October 31. By day five, eleven were gone. But something in how a couple of them disappeared bugged me. Our dedup logic only matches listings by source-and-listing-ID, never by VIN, and never across sources — on purpose, because two different sellers really can have the same VIN in a marketplace this size, and we never want to collapse two real ads into one. That rule exists for a good reason. It also means a dealer can pull an ad down and put the exact same car back up an hour later as a brand-new listing ID, and our system has no way of knowing it's the same metal. It just looks like a sale, then a new arrival.
So I went looking for the "sold" ones in their region the following day. Two of the eleven weren't sold at all. Same photos, same mileage, same VIN visible in the description — new listing ID, and a price exactly 400 AED higher than before. Not flagged as a deal anymore, because the cohort math had moved just enough.
That's the thing nobody tells you about a flagged deal: the badge is a snapshot of a price against a cohort at one moment. It isn't a verdict on the car, and it isn't permanent. The car didn't get worse. The arithmetic just reset.
November 2. A week in. Of the original forty, nineteen were still live with the badge intact. Twelve had vanished outright — genuinely gone, no relist I could find, which I'm reading as actual sales or withdrawals. Six had reappeared as the same car at a new price, like the Dubai sedan's cousin. Three I honestly couldn't resolve either way; the listing text was too thin to match confidently, which is its own small lesson about how much of this work is limited by how little some sellers write.
November 4. This is where it got genuinely interesting. Of the six relists, four came back cheaper than before, not pricier. A wagon out of Poland had been flagged, sat for nine days, got pulled, and came back 1,100 PLN lower — now flagged even harder. Whoever owns that listing is clearly watching the market the same way we are, just without our cohort math, and is cutting price in increments until it moves. That's not a broken system gaming our badge. That's just a seller slowly finding the floor, in public, in real time, and our data happens to be watching it happen.
November 6. By day eleven I'd lost two more to clean sales and gained one surprise: a Canadian listing that had been flagged, delisted, and came back with a price increase of nearly 9%. I checked the mileage field. It had dropped by around 3,000 km from the original post — a correction, not a new car. Someone had simply fat-fingered the odometer the first time, our model priced it against the wrong basket of comparable mileage, and the badge was arguably wrong from the start. That's a useful, slightly humbling data point: not every flagged deal that "stops being a deal" stopped being underpriced. Sometimes the input was bad and the badge corrects itself along with the listing.
November 8, closing the log. Final tally on the forty: fourteen sold or genuinely withdrawn, eight relisted at a lower price, four relisted higher (three of those for a documented reason — a correction, a trim detail added, a VAT line clarified — one I still can't explain), and the rest still sitting there, badge on, waiting.
So here's what I'd actually tell a buyer looking at a flagged car tonight. The badge is a real signal, not a trick — most of what disappeared off my list disappeared because it sold, and that's the base rate you should expect. But if you see a "sold" deal and go hunting for the car's twin a few days later, don't be shocked if you find it, slightly cheaper, slightly higher, or with one field quietly fixed. Across more than 1.1 million listings and 1,376 sources, a lot of what looks like churn is really the same seller testing the water in public. The badge tells you the price looked good the moment we checked. It was never going to promise you the moment would last.
