Find the previous task and the saved check from after the fix. Then compare the new problem URL with the one that was accepted at the time. The same finding can refer to a different page or a new version of a template.
For example, product names on regular product pages were fixed in September. In October, the finding appeared on a product page with an empty variant name. Start by comparing these two product pages, rather than asking for the previous fix to be repeated.
Check that it actually worked before
Start with the closed task. Were the test URLs, publication date, and result after publication saved? If acceptance consisted only of a message from the developer, there may be no verified state to compare against. Make this clear before comparing.
Check that the new report concerns the same URL. Pages may have similar names but different paths, parameters, or versions. One new product page with an issue does not prove that the problem has returned on the previously checked page.
Save the old successful result with its date. You need it for comparison, and a new failure does not remove it from the history. The process for checking the work is described in the acceptance guide.
Compare the conditions and pages
For a technical issue, the URL, page state, date, check method, and possible authentication matter. For template-related findings, check the page type and the content that affects the markup. First establish whether you are comparing the same things.
Check the audit coverage too. If the previous check covered one sample and the current one covers another, an increase in the number of affected pages may reflect broader coverage. The daily automated crawl is limited to 100 pages of the connected site, so an old finding being absent does not always mean it was absent from all the other URLs.
Also confirm when the change was published. The new report may have run before the change or while it was being rolled out. If the site serves different versions from cache, a manual check and an audit may disagree. Until the conditions match, do not call the difference a confirmed regression.
Send the developer a short history
| When | What was checked | Result |
|---|---|---|
| After the fix on 20 September | /product/1, regular product | The name is correct |
| 4 October | /product/2, empty variant field | The name is missing |
| 4 October | /product/1 checked again | The name is still correct |
Include another product with the same empty field, if there is one. Do not change several public-facing fields for an experiment without agreement. The developer can reproduce the case in a test environment.
Ask them to check how this data set is handled and whether it is related to the latest release. If the cause is unknown, say so. You can choose pages for checking the template based on their structure and content.
If the manual result and the audit disagree, include the time and conditions for each. A saved copy of the page, authentication, or a different URL may explain the difference, but the investigation—not a guess—should determine the explanation.
Check the new and previous cases after the fix
After publication, repeat the original check and investigate additional cases where the issue reappeared. For a template, this means at least the original URL, the new problematic example, and a suitable page that did not have the original issue.
Run a new technical report for the test URLs after publication. Save the results: the old product works, the new problematic case is fixed, and a similar case has also been checked. One successful run does not guarantee that the issue will never recur.
If the developer found the cause, add a specific rule for the next release—for example, check a product with an empty optional field after changing the template. Specify who will do the check and how in the task.
If the cause has not been established, leave the issue open and include any logs or release details needed. The issue disappearing temporarily after several actions does not show which action helped. Continue the task from a new observation time, keeping the previous evidence.