Skip to main content

How we verify findings and measure rule precision

We never invent a 0–100 compliance score. Every finding requires a verbatim quote verified against your fetched page HTML, and rules with fewer than 20 verified live samples display "Measuring" rather than estimated numbers.

1. Six finding states

Confirmed conflict

Two public surfaces on your store make contradictory statements under the same market and condition (backed by both quotes).

Missing information

A core policy link returns HTTP 404 or omits a required disclosure (such as return window or contact channel).

Please confirm

Worth checking against your actual business operations or Merchant Center settings; always shown free and never locked.

Consistent

Both compared surfaces were readable and matched.

Not checked

Page was behind a bot challenge, password, dynamic portal, or cross-currency view — never guessed as a conflict.

Fixed — verified

Awarded on re-check only when both previously conflicting pages are re-read live and now match.

2. Complete rule catalog & empirical precision ledger

Experimental rules (R16, R19, R20, R21) are mechanically capped at "Please confirm" until >=20 live external samples confirm >=95% precision.

CodeWhat we checkStageLive precisionSample nUpdated
R01
Shipping policy availability & delivery timeframes
Official reference
GAMeasuringMeasuring (<20)2026-11-22
R02
Return & refund policy availability & clarity
Official reference
GAMeasuringMeasuring (<20)2026-11-22
R03
Free-shipping threshold banner vs. shipping policy
Official reference
GAMeasuringMeasuring (<20)2026-11-22
R04
Return window days across policy, FAQ, and product tabs
Official reference
GAMeasuringMeasuring (<20)2026-11-22
R05
Who pays return shipping & restocking fees
Official reference
GAMeasuringMeasuring (<20)2026-11-22
R06
Visible product/variant price vs. structured data Google reads
Official reference
GAMeasuringMeasuring (<20)2026-11-22
R07
Unrecognized external or leftover template emails in policy text
Official reference
GAMeasuringMeasuring (<20)2026-11-22
R08
Customer contact channel availability & valid mailto links
Official reference
GAMeasuringMeasuring (<20)2026-11-22
R09
Unfilled theme placeholder text (Lorem ipsum / bracket tokens)
Official reference
GAMeasuringMeasuring (<20)2026-11-22
R10
Broken footer navigation links & empty social media icons
Official reference
GAMeasuringMeasuring (<20)2026-11-22
R11
Storefront robots.txt blocking Googlebot or Storebot-Google
Official reference
GAMeasuringMeasuring (<20)2026-11-22
R12
Cross-border entity registration vs. warehouse origin disclosure
Official reference
BetaMeasuringMeasuring (<20)2026-11-22
R13
Product images below 500×500 pixels (Jan 31, 2027 spec)
Official reference
GAMeasuringMeasuring (<20)2026-11-22
R14
Subscription, bundle, or currency-switcher apps altering displayed price
Official reference
BetaMeasuringMeasuring (<20)2026-11-22
R15
Unsubstantiated medical or cure claims in product copy
Official reference
BetaMeasuringMeasuring (<20)2026-11-22
R16
Extra paid items automatically added to cart (e.g. shipping protection)
Official reference
Experimental (capped)MeasuringMeasuring (<20)2026-11-22
R17
Resetting countdown timers, unlinked press logos, or supplier image hosts
Official reference
BetaMeasuringMeasuring (<20)2026-11-22
R18
UK & EU mandatory 14-day statutory return minimum
Official reference
BetaMeasuringMeasuring (<20)2026-11-22
R19
Street address consistency across pages & shared registered-agent suites
Official reference
Experimental (capped)MeasuringMeasuring (<20)2026-11-22
R20
Support email mail-server reachability & free-webmail check
Official reference
Experimental (capped)MeasuringMeasuring (<20)2026-11-22
R21
Domain registration age vs. "established since" claims on page
Official reference
Experimental (capped)MeasuringMeasuring (<20)2026-11-22
R22
Promotional text in product titles, catalog-wide strikethroughs & 0g weights
Official reference
BetaMeasuringMeasuring (<20)2026-11-22

3. How we test & calibrate rules

Golden test stores: Every code change runs against 12 purpose-built golden test stores (including 4 clean control stores) enforcing 0 false conflicts on clean stores and >=95% recall.

Heterogeneous dual-judge & human audit: Findings pass a mechanical verbatim quote gate, cross-family dual-judge verification, and merchant "This looks wrong" dispute reviews.

4. What public scans cannot see

  • Private identity documents, incorporation reviews, and linked-account history inside Merchant Center
  • Google Ads billing holds and payment profile verification
  • Password-locked storefronts or final carrier tax calculations requiring shopper PII

5. Methodology changelog

2026-11-22 — Capped experimental rules R16, R19, R20, and R21 at "Please confirm" pending >=20 live samples; shipped 5,000-SKU paginated binary image header inspection.