GetCarfeinatedPowered byCARFEINE
Claim a rooftopScan a dealership
Methodology · v1.3.0 · Updated 04 / 2026

How we score a dealership.

We grade every rooftop on the same rubric: six dimensions, twenty-seven evidence-backed checks, calibrated against the broader dealer cohort. Plain-spoken, no black box, no surprises.

27checks across
6dimensions, rolling up to
100composite points
Scan a dealershipSee the dimensions ↓
The six dimensions

Six dimensions. One hundred points.

Each dimension covers a distinct surface an AI buyer must verify before transacting. The rubric is calibrated quarterly against the broader dealer cohort, so the bar tracks the industry — not a static checklist.

Each dimension contributes to the composite · the rubric's internal weighting is part of Carfeine's calibrated scoring method.

Every check, named

The 27 checks at a glance.

Each check returns a pass / partial / fail with a snippet of evidence from your live site. We don't grade what we can't show you.

01

Inventory Intelligence

5 checks
  • INV-01
    Structured vehicle recordsEach VDP exposes machine-readable car details.
  • INV-02
    VIN coverageVIN is present, valid, and reachable per vehicle.
  • INV-03
    Photo + media depthAt least one canonical photo per car, alt-text resolvable.
  • INV-04
    Stock + status freshnessOn-lot vs. in-transit reflected within 24h.
  • INV-05
    Search + filter accessibilitySRP filters reachable without JavaScript.
02

Cost Transparency

5 checks
  • COST-01
    Out-the-door pricingTotal surfaces near the headline price, not in fine print.
  • COST-02
    Fees disclosedDoc fee, dealer adds, accessory fees enumerable.
  • COST-03
    Tax estimationLocal tax burden surfaced or estimable at quote time.
  • COST-04
    Rebate eligibilityConditional rebates marked as conditional.
  • COST-05
    Finance termsAPR / term assumptions stated alongside payments.
03

Compliance Readiness

4 checks
  • CMP-01
    Jurisdiction identificationDisclosures match the dealer's actual jurisdiction.
  • CMP-02
    Disclaimer presenceRequired regulatory disclaimers on the page.
  • CMP-03
    Price advertising signalsConditional offers and "with approved credit" surfaced clearly.
  • CMP-04
    Compliance metadataPII handling and tracking consent are declared.
04

Data Quality

4 checks
  • QLT-01
    Field completenessRequired attributes populated across the catalog.
  • QLT-02
    Accuracy signalsPrice and spec match between SRP, VDP, and feed.
  • QLT-03
    Freshness indicatorsRecords expose when they were last updated.
  • QLT-04
    Provenance signalsSource of record is identifiable per row.
05

Agent Tool Surface

5 checks
  • AGT-01
    Server-side MCPPublic Model Context Protocol surface an agent can call.
  • AGT-02
    WebMCP presenceBrowser-callable tool definitions for in-page agents.
  • AGT-03
    Tool manifest qualityAction surface documented: quote, hold, schedule — not just read.
  • AGT-04
    Authentication supportPredictable auth contract agents can plan against.
  • AGT-05
    Infrastructure baselinep95 latency + uptime within an agent's working budget.
06

Network Intelligence

4 checks
  • NET-01
    Network membershipInventory reachable through major aggregators.
  • NET-02
    Cross-dealer discoveryFindable via cohort indexes and rooftop directories.
  • NET-03
    Market positioningCohort-aware metadata exposed for placement.
  • NET-04
    Enrichment depthOpen feeds, partner integrations, reputation signals.
Anatomy of a scan

From a URL to a verdict in under a minute.

Five stages, every time. Same rubric for every rooftop, whether you're a 3-store group or a 130-store group.

  1. 01

    URL

    You drop a homepage. We accept just the domain — no install, no tags, no signup.

    savannah-toyota.com
  2. 02

    Crawl

    We read the homepage plus two to three key pages — a VDP, an SRP, a finance page — the way an agent would.

    homepage · /inventory · /vehicle/…
  3. 03

    Check

    Twenty-seven checks run against the captured pages. Each is a yes-and-how-much question, not a vibe.

    27 checks · 6 dimensions
  4. 04

    Evidence

    Every check returns a snippet — a tag, a value, a missing field — so you can see what we saw.

    VIN field present · 482 / 482 records
  5. 05

    Score

    Checks roll up into dimension scores; dimensions roll up into a 0–100 composite and a fuel level.

    62 / 100 · Half tank
Fuel levels

From Fumes to Supercharged.

Composite scores translate to a fuel level so a board, a GM, and an OEM rep all read the same answer the same way.

0–15

Fumes

An agent can neither find nor transact with this rooftop.

16–35

Reserve

Some data accessible. Agents can't complete a sale.

36–55

Quarter

Agents can read, but the work to transact hasn't been done.

56–75

Half

Agents can transact, with friction.

76–90

Full

Production-grade for AI commerce.

91–100

Supercharged

Reference implementation. Top 1%.

Calibration & freshness

Built to stay honest as the industry moves.

Cohort-calibrated

Every score is read against your cohort — same region, same OEM mix. As the industry improves, so does the bar.

Quarterly re-scans

Public scores re-run every quarter. Claimed rooftops can re-scan on demand after they ship a fix.

Rubric versioning

When we add or retire a check, the rubric version bumps and your prior scores stay viewable for trend integrity.

No black box

Every point traces back to evidence on your site. If we can't show you the snippet, we don't score it.

Want to see the rubric in action?

Scan a rooftop and see all twenty-seven checks resolved against a real site.

Scan your rooftopBrowse the leaderboard