Tony Wang12 min readWe Checked 133,052 Used-Car Listings. Autotrader Can't Tell AWD From 4WD for 64% of Its Own Inventory.
CarMax, Autotrader, and Cars.com use different field vocabularies — one flips a brand ranking, another hides two-thirds of a marketplace's drivetrain data.
Three sites, three data formats, one dataset. When Crawlora's Vehicle Listings API merges CarMax, Autotrader, and Cars.com into a single schema, most fields line up cleanly. A few don't — and the ones that don't aren't cosmetic. One flips which brand outranks which. Another means a marketplace's own drivetrain field is structurally incapable of answering "is this AWD or 4WD" for nearly two-thirds of its inventory.
This is a data-quality audit first and a market-numbers piece second: what's fragmented, what we merged, what we deliberately left alone, and what the corrected numbers say once the fragmentation is accounted for.
The fragmentation isn't random — it's one clean vocabulary per site
Every field that's split isn't split randomly across records. It's split by source: each site's own markup uses its own convention, and Crawlora's schema passes that convention through rather than silently guessing at a merge. That's the right default — but it means three specific fields need a manual reconciliation before you rank anything from them.
Make has exactly 5 case-duplicate pairs: RAM/Ram, MAZDA/Mazda, INFINITI/Infiniti, MINI/Mini, FIAT/Fiat. State splits by full name vs. abbreviation — CarMax emits California, Texas; Autotrader and Cars.com both emit CA, TX. Drive type splits into three separate per-source vocabularies entirely (more on that below).
For RAM and Mazda, merging the pairs barely moves anything — both were already correctly ranked around 14th and 16th on their larger raw bucket alone. For Infiniti and Mini, it's the difference between "outside the top 30" and "ahead of four recognizable competitor brands." Infiniti's merged total (INFINITI 791 + Infiniti 412 = 1,203) passes Genesis, Porsche, Mitsubishi, and Land Rover — none of which show up if you only read its largest raw bucket. Mini's merged total (Mini 327 + MINI 317 = 644) passes Jaguar (354), which its larger bucket alone narrowly trails. A ranking built off the raw facet, uncorrected, quietly under-ranks exactly the two brands whose listings happen to be split closest to a real threshold.
State has the same mechanism, and it's a bigger miss in absolute terms — not because it reorders anything, but because of how much of the two largest markets it hides.
Show the corrected top-10 states
| State | Naive (CarMax bucket only) | Corrected total |
|---|---|---|
| California | 7,770 | 15,208 |
| Texas | 6,947 | 13,536 |
| Florida | — | 7,252 |
| North Carolina | — | 6,781 |
| Georgia | — | 5,012 |
| Ohio | — | 4,646 |
| Tennessee | — | 4,411 |
| Maryland | — | 3,602 |
| Virginia | — | 3,530 |
| Pennsylvania | — | 3,389 |
California and Texas were already #1 and #2 either way — merging state formats doesn't flip a rank the way the make fix did. What it changes is magnitude: California's naive top bucket (California, CarMax only) is 7,770; the corrected total, adding CA from Autotrader and Cars.com, is 15,208 — essentially double. Texas goes from a naive 6,947 to a corrected 13,536. If you're citing "how many listings are in California" from this dataset, reading only one source's naming convention understates it by about half.
Autotrader's own drive-type field can't tell AWD from 4WD
This is the sharpest data-quality finding in the dataset, and it isn't a Crawlora aggregation artifact — it's the site's own field. CarMax and Cars.com each cleanly split all-wheel drive from four-wheel drive in their own listing data:
- CarMax:
Front Wheel Drive(21,981),All Wheel Drive(18,871),Four Wheel Drive(7,079),Rear Wheel Drive(5,098) — four clean values, no hyphens, zero cross-contamination. - Cars.com:
All-wheel Drive(19,831),Four-wheel Drive(18,740),Front-wheel Drive(8,246),Rear-wheel Drive(6,057) — hyphenated, plus a small leak (~195 records, under 0.4% of its own inventory) of the other two sources' abbreviated formats. - Autotrader:
AWD4WD(17,013),FWD(6,715),RWD(2,923) — only three buckets, ever. Filtering the facet tosource=autotraderalone still returns nothing but those three values.
17,013 of Autotrader's 26,632 listings — 63.9% of one of the three biggest used-car marketplaces — carry a value that simply doesn't say which drivetrain the vehicle has. Most people assume "AWD" and "4WD" are two different, clearly labeled things on any car-shopping site. For nearly two-thirds of Autotrader's own inventory, from this field alone, they aren't. For the 87.2% of the full dataset where the distinction does exist, the corrected split is FWD 36,971 (31.9%), AWD 38,744 (33.4%), 4WD 25,844 (22.3%), RWD 14,089 (12.1%).
CarMax and the marketplaces don't just have different prices — they have different shapes
Filtering each source by price band shows this isn't a subtle skew. It's structural, and it traces directly back to the sampling-method difference noted at the top: CarMax is one national retailer's enumerable inventory, while Autotrader and Cars.com are marketplace sweeps that pick up whatever independent dealers list — including luxury and exotic inventory a single used-car retailer doesn't carry.
CarMax is 63.7% under $30K (33,861 of 53,123) and only 1.7% is $50K+ (888 of 53,123). Cars.com is close to the inverse: 27.5% under $30K (14,636 of 53,297) and 37.4% at $50K+ (19,950 of 53,297). This is exactly why the dataset's own landing page shows Bentleys, Rolls-Royces, and a $480K Ferrari as its notable listings, and every one of those records is a Cars.com listing. It isn't cherry-picking a marketplace to make a point — it's that a marketplace-aggregator model structurally pulls in inventory a single national retailer's buying model never will.
Autotrader's "price reduced" flag isn't a marketplace behavior — it's a feature only Autotrader ships
Read on its own, "0% price-reduced on CarMax and Cars.com" could sound like those sellers never mark anything down. That's not the right read. The stable ~38–40% rate scaling cleanly with Autotrader's own crawl size is strong evidence this is a badge Autotrader's site exposes in its listing data and the other two simply don't expose at all — a field-coverage gap between sources, not a real difference in how often each marketplace's sellers discount.
EV and hybrid share looks flat — until you filter to what's actually new
Overall EV+hybrid share across the whole dataset holds at 9.3% (12,386 of 133,052: Electric 4,722, Hybrid 5,044, Hybrid Gas/Electric 1,434, Plug-In Hybrid 1,186) — almost unchanged from a prior pass's 9.2%. Read in isolation, a flat number like that could look like electrification has stalled in the used-car market. It hasn't; it's diluted by how many older gas vehicles simply exist and get resold.
12,386 of 133,052
9,587 of 80,200
Restricting to model year 2023 or newer — 80,200 of the 133,052 listings, 60.3% of the whole dataset — the EV+hybrid share rises to 12.0% (9,587 of 80,200: Electric 3,782, Hybrid 3,903, Hybrid Gas/Electric 1,046, Plug-In Hybrid 856). The 2023+ slice is a much better proxy for what's actually being built and sold today than the headline number spanning every model year on the lot, and by that measure electrification is running about 30% higher than the flat overall figure suggests.
What we didn't chase down
body_style is fragmented by the same per-source-vocabulary mechanism as drive_type and state — CarMax uses door-count-prefixed forms (4D Sport Utility, 4D Sedan, 2D Coupe); Cars.com and Autotrader use plain category names (SUV, Sedan, Coupe). Unlike drive_type's clean four-bucket target, body style has genuinely ambiguous edges — does a 4D Crew Cab belong with Truck, or is it its own category? — that need a real mapping-table decision this piece isn't making. fuel_type has minor sub-20-record fragmentation beyond the known Gasoline/Gas pair (a Flexible Fuel vs. E85 Flex Fuel split, and one literal truncation bug), left unmerged because none of it moves any published percentage by more than 0.02 points. And VIN coverage was spot-checked at 30 of 30 in the newest-listings slice only — there's no has_vin filter to verify completeness across the full ~133K, so we're not citing an overall VIN rate.
All figures above are a snapshot from 2026-08-26; the dataset visibly grew by roughly 150 records during the research session itself, consistent with its normal growth rate, so small residual drift (tens to low hundreds of records) is expected in any long-tail rank near a facet's aggregation cutoff.
What it's for
CarMax, Autotrader, and Cars.com listings — make, model, price, drive type, body style, and more — are queryable at /datasets/vehicle-listings over one REST API, normalized into a single schema. Normalized doesn't mean pre-reconciled: as shown above, a handful of fields still carry each source's own vocabulary through to the API, and a ranking built without accounting for that will quietly misstate exactly the brands, states, and drivetrains sitting closest to a merge boundary.
Sources
Related reading
Frequently asked questions
Can Autotrader's own data tell whether a car is all-wheel drive or four-wheel drive?
Not for most of its inventory. 17,013 of Autotrader's 26,632 listings (63.9%) carry a single combined drive_type value, AWD4WD, with no way to split it further. Filtering the facet to source=autotrader alone still returns only AWD4WD/FWD/RWD — three buckets, never a clean AWD/4WD split — confirming this is a limitation of Autotrader's own listing data, not a Crawlora aggregation issue. CarMax and Cars.com both cleanly separate the two in their own data.
Why does merging case-duplicate car brand names (like INFINITI vs Infiniti) matter?
Because it changes rank order, not just totals. Crawlora's Vehicle Listings dataset has 5 known case-duplicate make pairs (RAM/Ram, MAZDA/Mazda, INFINITI/Infiniti, MINI/Mini, FIAT/Fiat). For RAM and Mazda, merging barely moves their position. For Infiniti, merging INFINITI (791) and Infiniti (412) into 1,203 moves it from outside the top 30 to #26 overall, ahead of Genesis, Porsche, Mitsubishi, and Land Rover — none of which are visible if you read only its larger raw bucket.
Do CarMax, Autotrader, and Cars.com list the same mix of used cars?
No — the price distributions are close to mirror images. CarMax is 63.7% under $30,000 and only 1.7% is $50,000+. Cars.com inverts that: 27.5% under $30,000, 37.4% at $50,000+. This traces to a real sampling difference: CarMax rows are one national retailer's enumerable inventory, while Autotrader and Cars.com are marketplace sweeps that also pick up independent luxury and exotic dealer listings.
Is Autotrader's 'price reduced' flag evidence that only Autotrader sellers discount?
No. is_price_reduced returns exactly 0 for both CarMax and Cars.com and 39.9% for Autotrader (10,635 of 26,632) — a rate that held steady even as Autotrader's own crawl coverage roughly doubled between two passes. That stability is consistent with it being a badge only Autotrader's site exposes in its data, not a genuine difference in how often each marketplace's sellers actually discount.
Has EV and hybrid adoption stalled in the used-car market?
Not among what's actually newer. Overall EV+hybrid share is a flat 9.3% of all 133,052 listings, which could read as a plateau. But restricted to 2023-or-newer model years (60.3% of the dataset), the share rises to 12.0% — about 30% higher — because the overall figure is diluted by the much larger pool of older gas vehicles still being resold.
How much bigger are California's and Texas's true used-car listing counts than they first appear?
About double. CarMax reports state as a full name (California, Texas); Autotrader and Cars.com both report a two-letter abbreviation (CA, TX) that never merges automatically. California's naive count reading only CarMax's bucket is 7,770; the corrected total adding the CA abbreviation from the other two sources is 15,208. Texas goes from a naive 6,947 to a corrected 13,536.