What 30,000 Listings and 800 Sales Show About Marketplace
29,969 vehicles imported and 802 sold units, measured 25 August 2026. The median unit sold 18.0 days after its most recent post, with 65.2% inside 30 days. Photo count did not predict speed. Price band barely did. What predicted it was who posted the unit — the gap between the fastest and slowest dealer's median was 46.9 days, roughly eight times the spread across every price band combined.
Nobody in this market publishes sold-volume data. The numbers competitors put on their homepages are static claims — "73+ dealerships", "4,000+ users", "4,500+", "8,000+ members" — and two of them animate hard-coded figures up from zero on page load. None of it is checkable and none of it tells a dealer anything about what happens after a listing goes up.
This is our attempt at the other thing. Every figure below comes with its sample size, its method, and the reasons it might be wrong. Where an analysis failed, it is reported as failed.
Method
Two datasets, both from autobook.io's production database, both United States dealer inventory.
Import quality. Every vehicle imported in a trailing 90-day window, n = 29,969. Per-field landing rates measured on the record as created, before any human editing.
Time to sale. Every unit with a sale attributed through Smart Sale Detection, n = 802. The clock starts at the vehicle's most recent post and ends at the sale event.
Both are dealer inventory posted through one platform, which is a real limitation. For context on the wider market, the US Census Bureau publishes monthly retail trade time series for motor vehicle and parts dealers, including the seasonal factors we use elsewhere. This is not a Marketplace-wide sample and nothing here should be read as describing the platform as a whole. It describes what happens to dealer units posted by people paying for software to post them, which is a narrower and probably better-run population than Marketplace at large.
The correction that produced these numbers
This section exists because the alternative is quietly restating a number that changed.
An earlier version of this analysis measured the sale as the last time anything modified the vehicle's record. That is wrong, and it took a while to notice why: the field moves whenever anything writes to the row, so an unrelated bulk operation can rewrite a sale date months after the sale happened. In August 2026 exactly that occurred to several hundred rows, and the published median moved from 24 days to 50.5 overnight without a single car changing hands.
The figures here come from the sale event itself — a record written once when the sale is detected and never edited afterwards. Re-run on that basis, the median is 18.0 days. The previously published 24.0 is withdrawn.
We checked the new source for the same defect. It has its own clustering — 467 sale events share one date, presumably a detection backfill — and zero of those rows survive the join, because they belong to vehicles with no post timestamp or a sale predating it. The 802-unit sample contains none of them.
The rule that follows, for anyone doing this on their own data: never build a published metric on a column that anything else is allowed to write to. If a figure cannot be reproduced from an append-only event, it is not a measurement.
Finding 1 — What arrives when you import a dealership website
n = 29,969 vehicles, trailing 90 days.
| Field | Lands on |
|---|---|
| Description present | 100.0% |
| Transmission | 99.8% |
| Exterior colour | 99.8% |
| VIN (17 characters) | 99.5% |
| Mileage | 98.0% |
| Trim | 97.9% |
| A real price (excludes $0 and $1 placeholders) | 97.8% |
| At least one photo | 94.6% |
| Median photos per vehicle | 20 |
| Arriving at Facebook's 20-photo cap | 55.9% |
The miss is the interesting number. 5.4% of imported units arrive with no photo at all — roughly one in nineteen — and a vehicle listing with no photos will not sell. 2.2% arrive without a usable price, almost always because the source website shows "call for price" and the field resolves to a placeholder.
The photo median deserves reading carefully. Twenty is not "about twenty"; it is the ceiling. Facebook caps a vehicle listing at 20 photos, more than half of all units land exactly on it, and the distribution does not continue past it. Any claim you see about listings with more than twenty photos is describing something that cannot exist.
Finding 2 — How long a unit takes to sell
n = 802 sold units, measured 25 August 2026.
| Measure | Days from most recent post |
|---|---|
| Sold within 24 hours | 2.9% of units |
| Sold within 7 days | 18.3% |
| Fastest quarter (p25) | 9.1 days or less |
| Median | 18.0 days |
| Mean | 28.5 days |
| Sold within 30 days | 65.2% |
| Slowest quarter (p75) | 44.1 days or more |
| Longest observed | 152 days |
Every figure here is a floor. The clock starts at the unit's most recent post, because that is what the record stores. A repost resets it. So a unit listed on 1 May, relisted on 1 June and sold on 25 June reads as 24 days in our data, and the rep lived through 55.
A second bias runs the other way. Sale times are recorded when the sale is detected, and plenty of reps clear their board in one sitting, marking a run of units sold in the same minute. That drags the measured median later than the truth. One bias makes 18 days look short and the other makes it look long, and we are not going to pretend they cancel neatly.
The mean sitting ten days above the median is the shape of a long tail: most units go reasonably quickly and a minority sit for months.
Finding 3 — Photo count did not predict speed
We expected this to be the headline. It is not, and the null result is more useful than the finding we wanted.
| Photos on the listing | Units | Median days | Sold within 30 days |
|---|---|---|---|
| 1–5 | 61 | 19.9 | 67.2% |
| 6–10 | 22 | 19.0 | 63.6% |
| 11–19 | 374 | 18.9 | 64.4% |
| 20 (the cap) | 345 | 15.9 | 65.8% |
The 30-day column is flat. Every bucket lands between 63.6% and 67.2%, including the one with five photos or fewer. The median moves about three days in favour of listings at the cap, and the 6–10 bucket has 22 units in it, which is not enough to lean on.
We looked for "more photos sell cars faster" and did not find it. Anyone publishing a figure like "listings with 15+ photos sell 40% faster" is either sitting on a controlled study they have not shown you or making it up.
This does not mean photo count is irrelevant. It means it does not measurably move time to sale in this sample. The argument for filling all twenty is different and still holds: every photo is a question a buyer does not have to ask, and buyer messages are the expensive part of this channel.
Finding 4 — Price band moves it, barely
| Price band | Units | Median days | Sold within 30 days |
|---|---|---|---|
| Under $10,000 | 50 | 13.4 | 78.0% |
| $10,000–20,000 | 151 | 16.2 | 69.5% |
| $20,000–30,000 | 194 | 18.5 | 59.8% |
| $30,000 and up | 406 | 18.9 | 64.8% |
The median rises monotonically with price, which is the direction everyone expects. The size is the surprise: 5.5 days across the entire range.
And the 30-day column misbehaves — the $30,000-plus band beats the $20,000–30,000 band, which it should not if price alone drove this. The under-$10,000 band is also only 50 units.
So the defensible statement is narrow: cheaper units sell somewhat faster, the effect is smaller than the industry assumes, and it is not clean enough to plan around on its own.
Finding 5 — Who posts the unit is the whole story
This is the finding that matters, and it is the one nobody sells software against.
Taking every dealer with at least 10 sold units in the sample (14 dealers) and computing each one's median:
| Measure | Value |
|---|---|
| Fastest dealer's median | 4.9 days |
| Median dealer's median | 20.8 days |
| Slowest dealer's median | 51.8 days |
| Dealer-to-dealer spread | 46.9 days |
| Price-band spread, for comparison | 5.5 days |
| Photo-count spread, for comparison | 4.0 days |
The gap between the best and worst operator is roughly eight and a half times the gap across every price band, and nearly twelve times the gap across photo counts.
Put plainly: the same vehicle, at the same price, with the same number of photos, sells in a fifth of the time at one store versus another. Whatever is happening in that 46.9-day gap is worth more than every listing optimisation in this article combined.
We cannot tell you from this data exactly what it is, and we are not going to guess dressed up as analysis. What we can say is what it is not: it is not the price band, and it is not the photo count, because both were measured and both are small.
What we could not measure
Published so nobody assumes we simply omitted the inconvenient parts.
- Time of day and day of week. Permanently unavailable. The post timestamp records the most recent repost, so any hour-of-day analysis is corrupted at source. We cannot support any "best time to post" claim from our own data and have said so in the article about it.
- Body style. Not run this cycle. No body-style claims should be attributed to us.
- Conversion rate per listing. This is a study of units that sold. It says how long a sale took, not the odds of getting one. Across the whole network roughly one in nine posted vehicles gets an attributed sale, and that counter updates automatically on our comparison pages.
- Anything about Marketplace generally. One platform's dealer inventory. Not a census.
- Description length, listing title wording, renewal effects. Not measured. Any advice we give on those is reasoning, not evidence, and is labelled that way.
What a rep should take from this
Four things, in order of how much they matter.
Day seven is not a signal. Only 18.3% of units sold inside a week. A quiet first week is the normal shape of this channel, and cutting a price in response to it is reacting to noise. The point where the number genuinely deserves a second look is around day 18, and past roughly 44 days the unit is in the slowest quarter — the full decision rule is in when to lower your price, and the 20-photo cap this study measures against is Facebook's, set out in their Commerce Policies.
Check the 5.4%. Whatever tool or process you use, some share of your inventory reaches Marketplace without photos. Ours is one in nineteen. Find yours, because those listings will not sell and nothing downstream fixes them — the usual causes are in how many photos a listing needs.
Fill the twenty, for the right reason. Not because it sells faster — we looked and it does not, measurably. Because every photo retires a buyer question, and buyer questions cost about forty minutes a day, which is what lowballers, no-shows and time wasters is about.
The operator gap is where the money is. A 46.9-day spread between dealers dwarfs everything else measured here. Consistency of posting, speed of reply and how the listings are maintained are not glamorous variables, and they appear to matter more than any attribute of the vehicle.
Reuse and citation
Every figure here may be quoted with attribution to autobook.io and the measurement date of 25 August 2026. If you cite the time-to-sale numbers, please carry the floor caveat with them — the clock starts at the most recent post, so they understate true time from first listing.
We intend to re-run this quarterly and to publish it when the numbers move against us, which is the only version of this exercise that is worth anything. The correction at the top of this page is the first instance of that.
Frequently asked questions
How long does it take to sell a car on Facebook Marketplace?
In this sample of 802 dealer units, a median of 18.0 days from the most recent post, with 65.2% sold inside 30 days and a slowest quarter running past 44 days. Read every figure as a floor.
Do more photos sell cars faster on Marketplace?
We could not find that effect. The 30-day sold share was flat across every photo bucket from 1–5 up to the 20-photo cap.
Do cheaper cars sell faster?
Somewhat. Median days rose with price across all four bands, but the total spread was 5.5 days and the 30-day share was not monotonic.
What actually predicts how fast a unit sells?
The dealer. The spread between the fastest and slowest operator's median was 46.9 days, about eight and a half times the spread across price bands.
Why did your median change from 24 days to 18?
The old figure was measured from a database column that unrelated updates could rewrite. These figures come from the sale event, which is written once. The old number is withdrawn.
Is this Marketplace-wide data?
No. It is United States dealer inventory posted through one platform — a narrower and probably better-run population than Marketplace as a whole.