Resolution Layer Case Study · Read & Diagnose
A National Franchise Dealer Group
A cross-rooftop new-vehicle catalog analysis — 5 stores, 4 brands, 2 platforms, zero manual setup
Before the findings
What's actually riding on five rooftops' worth of vehicle data.
About this case study. This is an anonymized version of a client deliverable. The dealer group, its rooftops, its locations and its history have been removed; every finding, severity, count, field name and verbatim data value is reproduced exactly as it appeared in the original analysis. Rooftops are labeled by number and franchise brand only.
This group's reputation was built one rooftop at a time — dozens of dealerships across more than a dozen states, each store run by a team that knows its own lot, its own local market, its own customers. That reputation is the asset. What this analysis examines is the catalog infrastructure that either supports it or quietly works against it.
Every rooftop, regardless of brand or platform, feeds the same handful of channels that decide whether a shopper ever sees a given vehicle: the dealership's own site, the big three marketplaces — Autotrader, Cars.com, CarGurus — and, increasingly, a channel most dealer teams haven't built for at all.
"By bringing CarMax's nationwide inventory and shopping tools into ChatGPT, we're giving customers a simple, reliable way to explore their options within a platform they are already using."
Diane Cafritz, Chief Innovation & People Officer, CarMax — press release, Feb. 27, 2026
We ran five of the group's rooftops — two Chevrolet stores, one Honda, one Kia, one Volkswagen, across two different inventory platforms — through EKOM's analysis pipeline exactly as we would a new customer's catalog. No hand-built schema, no per-store configuration, no one telling the system what to look for. What came back wasn't five stores' worth of unrelated mistakes. It was the same handful of defect classes, surfacing independently, store after store.
At a glance
A machine-surfaced triage signal, not a verified defect list — the vast majority are real; a small share may reflect intentional dealer choices. Every item below is worth a human look.
What recurred everywhere
Both Chevrolet stores · 140+ vehicles
4WD trucks tagged "AWD"
Four-Wheel Drive and All-Wheel Drive are different systems, sold differently and searched differently. Store 1 and Store 2 made the identical mistake, independently — a shopper filtering for AWD sees trucks that aren't AWD; a shopper filtering for 4WD misses trucks that are.
Chevrolet, Chevrolet, Volkswagen · 3 rooftops
Fuel type left blank — including on EVs
Store 3's fuel_type field is blank for every vehicle in the catalog, including electric ID.4 models. An EV that can't be found by a shopper filtering for electric isn't a data nit — it's a lead the store never gets a chance at.
Chevrolet AND Honda · different brand, same mistake
Pre-owned language on new-vehicle listings
New-inventory descriptions at both Store 1 (Chevrolet) and Store 5 (Honda) carry boilerplate stating the vehicle underwent a "rigorous inspection process" for pre-owned sale. On a new vehicle, that's not a copy error — it's a disclosure/compliance exposure, and it appears on two unrelated brand platforms.
4 of 5 rooftops
Interior color contradicts the vehicle's own description
Store 4's K4s list "Black" interior while the dealer's own description says "Gray." Store 5 collapses Ridgeline TrailSport's signature orange stitching to plain "Black." The pattern — color fields disagreeing with the vehicle's own copy — shows up in some form at four of the five stores.
3 of 5 rooftops · up to 15% of a catalog
Missing MSRP or dealer price
Roughly 15% of Store 1's catalog is missing MSRP; Store 3 has ~55 vehicles with a price but no MSRP to show the savings against. A vehicle with no price is a vehicle a shopper can't sort, compare, or filter for on any platform.
Store 5 · Honda
One rooftop's branding bleeding into another's listings
A corporate group-level image served from a different dealership's subdomain shows up as the second photo on most fully-photographed Store 5 listings — a small, concrete signal that shared infrastructure across rooftops can carry a defect from one store into another's shopper experience.
By rooftop
Five independent runs, no cross-talk between them — the pattern above emerged without any store informing another.
| Rooftop | Brand | Platform | Findings | Fixes Proposed | High-Sev. |
|---|
| Store 1 | Chevrolet | Dealer Inspire | 89 | 568 | 77 |
| Store 2 | Chevrolet | Dealer Inspire | 87 | 118 | 69 |
| Store 3 | Volkswagen | Dealer Inspire | 64 | 226 | 45 |
| Store 4 | Kia | Dealer Inspire | 14 | 25 | 14 |
| Store 5 | Honda | Dealer.com | 12 | 69 | 11 |
| Total | 4 brands | 2 platforms | 266 | 1,006 | 216 |
A focused pass
A handful of findings are worth pulling out on their own — not because they're the most common, but because of what they reveal about how the data actually broke.
Critical · Store 2 · Chevrolet
A 2026 Tahoe Premier's engine field reads "<City>, <ST>" — verbatim, a dealership city-and-state string sitting where the powertrain belongs. It's a small error with a large implication: whatever pipeline maps data into that field isn't validating that what lands there is actually an engine.
Critical · Store 4 · Kia
A 2027 Telluride's engine field carries the Sportage Hybrid's 1.6L engine string instead of the Telluride's own 2.5L Turbo — spec data from one model bleeding into the listing for a completely different one.
Critical · Store 3 · Volkswagen
A Taos SE AWD is priced roughly $30,000 below its MSRP and every comparable sibling — appearing as the cheapest vehicle in the entire catalog by a factor of ten. Whether it's a feed error or a live pricing liability, it needs eyes today, not at the next data refresh.
Note on scope. Findings above the fold reflect the full run across all five rooftops. The complete list — 266 findings — is summarized by theme on the next page.
Findings by theme
Every finding maps to one of six recurring defect classes across the five rooftops.
| Finding | Sev. | Who feels it | Rooftops |
|---|
| Drivetrain & Powertrain |
| 4WD trucks tagged "AWD" in features | CRIT | Shoppers · Marketplaces | Both Chevrolet stores |
| Fuel type blank, incl. EVs and diesels | CRIT | Shoppers | Stores 1, 2, 3 |
| Fuel-economy strings stored in engine field | CRIT | Shoppers · Ops | Both Chevrolet stores |
| HR-V hybrids classified as "Gasoline" | CRIT | Shoppers | Store 5 (27 units) |
| Wrong-model engine spec pasted into listing | CRIT | Shoppers | Store 4 |
| Pricing & MSRP |
| Missing MSRP and/or dealer price | CRIT | Shoppers · Commerce | Stores 1, 2, 3 |
| Systematic pricing-feed artifacts (fixed-dollar deviations) | CRIT | Ops · Finance | Stores 1, 2 |
| MSRP/dealer-price outliers vs. cohort peers | HIGH | Shoppers · Finance | Stores 1, 2, 3 |
| Vehicle Identity (Stock Numbers) |
| Corrupted / malformed stock numbers | CRIT | Ops · DMS | Stores 1, 2, 3, 4 |
| Undocumented prefix schemes (4–5 variants per store) | HIGH | Ops | All 5 rooftops |
| Imagery |
| Thin (1–4 image) or generic placeholder photo sets | HIGH | Shoppers | All 5 rooftops |
| Shared / cross-vehicle images displayed as unit-specific | HIGH | Shoppers | Stores 1, 2, 3 |
| Cross-rooftop branding image bleed | MED | Shoppers | Store 5 |
| Content & Compliance |
| Pre-owned boilerplate on new-vehicle listings | CRIT | Compliance · Shoppers | Stores 1, 5 |
| Boilerplate-only descriptions, zero vehicle-specific content | HIGH | Shoppers · SEO | Stores 1, 2, 3, 4 |
| Structural (Compound Fields) |
| features_raw / image_urls / interior_color / dealer_description combine multiple data types in one field | HIGH | Ops · Engineering | All 5 rooftops |
Catalog structure
What's solid across all five rooftops, and where the schema itself is working against the data.
Strong coverage
- VIN 100%
- Make / Model / Year 100%
- Body Style ~99%
- Trim ~98%
Structural gaps
- fuel_type blank at 3 of 5
- mpg_city / mpg_highway no dedicated field
- condition absent — required by Google Vehicle Ads
- cab_style absent on truck inventory
Compound-split candidates (flagged across all 5 rooftops)
| Field | What it's carrying | Rooftops affected |
|---|
| features_raw | A flat, unparsed feature list — up to 200+ presence flags packed into one string | All 5 |
| image_urls | Photo URLs with no structured distinction between hero, interior, and stock/placeholder images | All 5 |
| interior_color | Color name + seat material combined in one free-text string | 4 of 5 |
| dealer_description | Marketing copy, transfer disclaimers, and compliance boilerplate concatenated with no field-level separation | 4 of 5 |
Note. These are flagged as split candidates, not pre-transformed — the underlying values are intentionally left packed so the recommendation stays visible rather than silently applied.
What this means
more than
25%
of new-car buyers now research using an AI tool like ChatGPT or Google AI Overviews before they ever contact a dealer — a channel the group's data is either visible to, or invisible to, with nothing in between.
Cox Automotive, 2026 Car Buyer Journey Study
The group's position in its markets was built on a promise of consistency — a customer who trusts one rooftop is meant to trust every one of them. Right now, five of those rooftops are each running an undetected version of the same handful of defects, independently, with no rooftop aware the others have the identical problem.
The commercial exposure is specific, not abstract: an all-electric VW lineup that's invisible to every shopper filtering for "electric." Chevrolet trucks that read as AWD when they're not, misdirecting exactly the buyers most likely to walk. A Taos priced ten times below every comparable unit. None of these cost a sale on their own — but each one is a lead that never converts, on a channel the group already paid to be listed on.
The pre-owned boilerplate finding deserves its own line, because it isn't a content problem — it's a disclosure problem, and it shows up independently on two different brand platforms. A new-vehicle listing that tells a shopper their car underwent a "rigorous pre-owned inspection process" is the kind of finding that matters to a compliance team, not just a merchandising one.
What EKOM brings is not just the capability to find what's wrong across five rooftops without being told where to look — it's the fact that the same pipeline runs the same way at rooftop six, sixty, and ninety, without asking any of them to change how they work. The pattern we found in five stores is very likely present across the rest of the group.
This is what one unattended pass surfaces with no access, no schema, and no one pointing at the problem.
The same pipeline runs the same way on the next rooftop, and the next platform.