A finding you can inspect.

Discount lost after a payment retry

Review required. The expected final total was €80; the observed total was €100 on both execution attempts.

Finding CHECKOUT-001 · Owned local fixture, build fixture-regression-v1 · Executed 30 September 2026, 05:07:50–05:07:58 UTC.

Reproduction steps

1. Add the €100 headphones to the controlled store’s cart.
2. Apply coupon SAVE20; confirm the displayed total is €80.
3. Select the simulated declined test-card action; confirm “Payment declined.”
4. Select the simulated valid-card retry action; confirm “Payment successful.”
5. Check that the discounted total remains €80.

Observed: the final total was €100. Both execution attempts failed the final €80 assertion. The payment retry above is part of the journey; the automated test retry reran the entire check once.

The saved report classifies this as suspected_product_defect. Repeated assertion failure is observed evidence; that label does not establish an automatically verified root cause.

Results and untested scope

Four distinct checks were selected: three browser checks and one API check. Three passed and one failed. There were five execution attempts: four initial attempts plus one automated retry. No selected checks were blocked or skipped. Account settings was not selected and remains untested.

What the evidence establishes

Two screenshots, two separate Playwright traces and two error-context records cover the failed attempts. Six additional log/response attachments cover the executed checks, including the passing quote API response.

All 12 artifact files matched the byte counts and SHA-256 fingerprints in the saved report when reviewed for this sample. The screenshot was visually inspected. Raw traces and logs are not published here. Fingerprint matching checks consistency with the saved report; it is not independent or signed attestation.

• This bounded four-check example is not full application coverage or a complete four-week pilot report.
• The change manifest, coverage and business-impact inputs are synthetic. No customer savings or defect probability is claimed.
• The recorded browser was system Chrome 154.0.8037.92 with Playwright 1.63.0. It was not a pinned browser image; video recording was disabled.
• This demonstrates local evidence collection, not hosted customer isolation, production readiness, security certification or automated release approval.