Catalog Intelligence · Three Groups, Independently Analyzed
Three dealer groups. Nineteen rooftops.
The same defects.
Nineteen rooftops across three unrelated franchise dealer groups, each read cold through the same unattended pipeline — no schema supplied, no per-store setup, nobody telling the system what to look for. The groups share no ownership, no market, and not always a platform. They produced the same handful of defect classes anyway.
6
Defect Classes
In All Three
The three groups
| Group | Franchise mix | Storefront | Rooftops | Findings | High-Sev |
|---|
| Group ANational, 90+ rooftops |
Chevrolet ×2, Honda, Kia, VW |
2 platforms |
5 | 266 | 216 · 81% |
| Group BLuxury, built by acquisition |
Mercedes-Benz ×6, Infiniti |
1 platform |
7 | 168 | 154 · 92% |
| Group CRegional, 13 franchises |
CDJR, VW, Acura, Audi, Infiniti, Porsche + 4 |
1 storefront |
7 | 143 | 119 · 83% |
| Combined | 20+ franchise brands | 2 platforms | 19 | 577 | 489 · 85% |
Six defect classes that surfaced in all three groups
Drivetrain labeled so shoppers can't filter it
4WD trucks tagged "AWD." One drivetrain written "4MATIC®" at some stores, "AWD 4MATIC®" at others. 4x4 / AWD / All-Wheel Drive used interchangeably in one feed. A shopper filtering for one never sees the other.
All 3 groups · 8 rooftops
Fuel type wrong, blank, or contradictory
One store's fuel_type blank for every vehicle, including its EVs. A G 63 AMG® listed with an "Electric Motor" — at two unrelated rooftops. Hybrids classified "Gasoline."
All 3 groups · 11 rooftops
Vehicles a price filter can't see
Missing MSRP on up to 15% of one catalog. Unpublished prices stored as msrp = "0", so savings math reads them as zero-dollar cars. A dealer price exactly $699 above MSRP.
All 3 groups · 10 rooftops
Condition stated wrong — a disclosure problem, not a copy one
Pre-owned inspection boilerplate on new-vehicle listings, on two unrelated brand platforms. A Certified Pre-Owned unit inside the new feed. "Used" cars on /certified/ URLs.
All 3 groups · compliance exposure
Spec data bleeding between models
A Telluride carrying a Sportage Hybrid's engine string. A Grand Cherokee described as a V6 that is an inline-4. An engine field whose value is a city and state.
All 3 groups
One field carrying four kinds of data
engine, interior_color, features_raw and dealer_description each pack several data types into one free-text string — at every rooftop in all three groups.
19 of 19 rooftops
already
44%
of car shoppers use an AI-powered search tool when researching a vehicle — and trust what it returns more than most dealers assume. A blank fuel type, an unparseable price, or a 4WD truck tagged AWD is invisible to that shopper twice over: once to the tool, once to the person reading its answer.
Cars.com, AI in Car Shopping Consumer Survey, Nov. 2025
Three groups that have never compared notes produced the same six defect classes independently. That is not a store problem — it is a category problem: findable without inside access, and fixable once per convention rather than once per car.
Method. A machine-surfaced triage signal, not a verified defect list — the vast majority are real; a small share may reflect intentional dealer choices. Anonymized — names, locations and histories removed; every finding, count, severity and verbatim value reproduced exactly as in the original client analyses.