The preview finally shows the correct heading. Before closing the issue, check the steps immediately around it: did the source save, and can the intended viewer see the result?
You do not have to test the whole application. Choose the few checks most likely to reveal a side effect of this particular fix. That is what a narrow regression check is for.
Keep the original problem in view
Write the failing step in one sentence. For a fictional document workflow: “The private preview showed an old heading after the editor saved a change.”
Then record the actual correction and the result of repeating that same harmless action. If a different account, workspace or document was used, state that difference. It may limit how much the result tells you about the original issue.
Do not replace the original symptom with a broader success claim such as “The publishing system works.” The fix has only established what the test actually observed.
Choose nearby checks the change could affect
For the fictional preview issue, relevant checks might include whether the source still saves, whether another paragraph remains intact and whether an authorized viewer sees the correct private draft.
An unrelated billing feature is not a useful neighbor. A live customer notification may be relevant to the wider workflow but still inappropriate to trigger without an isolated authorized test.
Choose neighbors based on the changed component and its dependencies. A shared page template has a wider potential effect than a personal display preference. If the change touched a global component, a single-page check is unlikely to establish its full scope.
Record the input, the fix and the next step
| Boundary | Expected behavior | Safe evidence | Result |
|---|---|---|---|
| Before the fixed step | Required input still saves correctly | Reopened private source | Actual observation |
| Fixed step | Original symptom no longer occurs | Same safe reproduction | Actual observation |
| After the fixed step | Intended role sees the right result | Authorized private view | Actual observation |
| Recovery | Prior state can be restored if needed | Verified supported recovery route | Established or unresolved |
Keep “not tested” as a valid result. A blank cell should not be interpreted as a pass, and a planned test should not appear as completed evidence.
Look beyond a reassuring preview
A preview can look correct while losing a link, alternate text or table column. For content changes, compare the saved source and the rendered result at the parts most likely to be affected.
For account configuration, reopen the saved field rather than relying solely on a confirmation toast. If there are two related scopes, verify that the intended one changed and the other retained its value.
This is especially important when a control’s wording does not make its scope obvious. Stop before saving if the change could affect more objects or people than authorized.
A useful closure might read: “The private preview now shows the saved heading. The source paragraph and authorized viewer check passed. Public publication and customer notifications were not tested.”
That statement tells the next person what is safe to rely on and what remains unknown. If a neighboring step fails, preserve the evidence and use the established recovery procedure where appropriate; do not keep adding unrelated fixes.
Fill the record for the component you changed. If the nearby steps pass, close that issue with the limits stated; if one fails, preserve the evidence and use the recovery plan where appropriate. Keep untested public actions or customer notifications separate. A clear statement of what works is more useful to the next person than “everything fixed.”