Here's an uncomfortable fact about the badge our own system prints on 12,823 listings a day, out of the 1.13 million we're currently tracking: it isn't a measurement of value. It's a measurement of how much comparison data existed at the moment that car got crawled. Most people read "below market" as "cheap for what it is." What it actually means, underneath, is "cheap relative to a cohort" — and cohorts are not created equal.
I build and run the thing that prints this badge, so let me walk through why that distinction matters more than it sounds.
How the flag actually gets decided
Deal flagging doesn't compare a car to some universal fair-value curve. It compares a listing to other listings like it — same region, roughly the same model and spec — and asks whether the asking price sits meaningfully below the pack. That's a sound approach when the pack is big. It gets shaky fast when the pack is small.
And most of our pack is small. The top 12 of our 73 tracked regions — UAE, France, Japan, South Korea, USA, Brazil, Czech Republic, Belgium, Canada, Poland, Portugal, Sweden — carry 548,120 listings between them, 48.5% of everything we have. The other 61 regions split the remaining 580,892 listings, which works out to an average of under 9,500 listings each. And that's before you slice by make and model. A region with 9,500 total listings spread across dozens of brands might have a few hundred of any given make, and a cohort in the low double digits for a specific model and trim once you get narrow enough.
Look at that last bar against the first twelve. That gap is the whole problem. A cohort built from 79,881 UAE listings has room to be statistically honest — outliers get drowned out, and a genuinely underpriced car stands out against real noise. A cohort built from a few dozen comparable cars in a thin region doesn't have that luxury. One oddly cheap trade-in, one seller who doesn't know the local market, one currency-conversion quirk, and the whole cohort's "below market" line moves.
A worked example, without inventing numbers I don't have
Take Volvo. We carry 23,742 Volvo listings globally — a real brand, not a rounding error, but spread across 73 regions rather than concentrated anywhere. In Sweden, Volvo's home market, that's a thick, well-behaved cohort: lots of comparable XC60s and V60s at every trim, so a flagged deal there is leaning on real statistical weight. Take that same model out to a region where Volvo barely registers, and the cohort the system is comparing against might be a handful of cars that happen to be for sale that month. The math still runs. The badge still prints. It just isn't carrying the same evidentiary weight it was a page ago.
| Thick cohort (e.g. a top-12 region, popular make) | Thin cohort (e.g. a tail region, niche make/trim) | |
|---|---|---|
| Comparable listings | Hundreds to thousands | Often single to low double digits |
| Effect of one outlier | Negligible | Can shift the whole "below market" line |
| Confidence in the flag | High | Low-to-moderate, even when the math is correct |
| What the badge is really telling you | This car is cheap relative to a real market | This car is cheap relative to whatever happened to be listed |
Neither row is "wrong." Both are the system doing exactly what it's built to do — flag anything that sits statistically below its peers, right now, in that place. The badge doesn't lie. It just doesn't come with an asterisk explaining its own sample size, and most readers assume every green tag was earned the same way.
Where the critics are right
And here's the part I'd concede if someone pushed back on this, because they'd have a point: a thin-cohort flag isn't worthless just because its sample is small. If a car really is priced well under everything else for sale in its own region — even if "everything else" is eleven cars, not eleven hundred — that's still real information a buyer wouldn't have had otherwise. Waiting for a thick sample before saying anything at all would mean staying silent in the 61 regions that need the signal most, since those are often the ones with the least pricing transparency to begin with. A thin flag in a data-poor market can be the only warning a buyer gets that a price is unusual, full stop.
So the honest position isn't "ignore flags from small regions." It's "treat the flag as a reason to look closer, not as a verdict already rendered." If a listing is flagged in a region that's one of our top 12, that's a strong, well-supported signal — lean on it. If it's flagged somewhere in that long tail of 61 regions averaging under 9,500 listings, it's a tip, not a conclusion. Pull the comparable listings yourself before you call the seller. The badge got you looking in the right direction. It didn't do the rest of the homework for you.
