A review that treats every finding as a stop sign cannot tell an owner whether the work is usable. A broken submit action, invisible keyboard focus and an uneven icon gap can all appear in the same report. They do not carry the same consequence. The useful question is what each finding prevents a person from doing, and what evidence supports that judgment.
I would agree on severity before asking an assistant to review the page. Otherwise, the review becomes an invitation to keep finding improvements until somebody loses patience. The assistant needs a rule for distinguishing a failed requirement from a preference, along with a person who can resolve the cases that resist classification. More comments do not make that decision easier by themselves.
Consider a hypothetical workshop registration page. The accepted task is to let a visitor enter valid details, submit once and receive confirmation, including when using a keyboard. The page must also preserve visible keyboard focus. Suppose a review reports three findings: submission does nothing, focused controls have no visible indicator, and the gap beside a decorative icon looks slightly uneven. These are invented observations for a paper exercise, not results from an application test.
The failed submit action blocks the central task. A useful report would identify the candidate build, supplied values, exact action and missing result. It would also distinguish a missing confirmation from a missing saved registration. Those can look identical to the visitor while requiring different repairs. A screenshot of the unchanged page establishes neither the storage outcome nor permission to submit repeatedly against a live service.
The focus finding requires repair too. W3C's explanation of Focus Visible (https://www.w3.org/WAI/WCAG22/Understanding/focus-visible.html) describes a Level AA requirement for a keyboard-operable interface to offer a mode in which the focus indicator is visible. That indicator tells a sighted keyboard user which control will receive an action. Calling it visual does not make it decorative; its appearance carries information needed to operate the page.
Keyboard operation has a separate requirement. W3C's Keyboard guidance (https://www.w3.org/WAI/WCAG22/Understanding/keyboard.html) explains that functionality must be available through a keyboard interface, subject to an exception for functions inherently dependent on a movement path. A registration form is not that exception. A visible outline alone does not prove submission works from the keyboard, and successful mouse submission does not prove the keyboard route works.
The icon gap is a different kind of finding only after its effect is understood. If it changes no meaning, obscures no content and violates no agreed acceptance condition, I would classify it as optional polish. If the spacing makes two controls difficult to distinguish or hides content at a supported size, the reviewer needs to report that consequence. The location of a defect in a stylesheet tells us little about its severity.
Discussion Topic:
Which causes more review friction on your team: unclear acceptance rules or cosmetic findings labeled as blockers?
A compact rubric can keep these distinctions usable. I would use four dispositions: blocker, required fix, optional polish and question. These are proposed workflow labels, not a universal industry scale or a classification prescribed by W3C. Their purpose is to tell the next person what must happen before acceptance, rather than leave them to infer urgency from an adjective in the review.
A blocker prevents the accepted task or leaves its result unsafe to accept. The broken submission belongs here. The report names the violated expectation, reproduction, impact and evidence on the candidate build. Work stays unaccepted until the failure is repaired and checked, or the owner explicitly changes the proposed release scope. A confident explanation of the likely cause cannot substitute for that retest.
A required fix violates an agreed requirement even when a narrower demonstration still works. In this example, missing visible focus belongs here: a mouse demonstration might succeed while the accepted keyboard experience remains defective. Required fixes also hold acceptance under this rubric. The distinction from a blocker helps describe impact and order the repair work; it does not create a back door for shipping a known accessibility failure.
Optional polish improves a result that already meets the agreed task. Record the suggestion and let the owner decide whether it deserves a separate change. The reviewer should be able to explain why deferring it preserves usability and acceptance. The uneven decorative gap can qualify on those terms. Calling something low effort is insufficient, since a tiny edit can still require verification and can introduce a new defect.
A question marks missing evidence or an unresolved rule. Perhaps the reviewer cannot determine whether confirmation must appear on the page or arrive by email. The owner must resolve that expectation before the reviewer can judge the behavior against it. A question about a required outcome can hold acceptance pending an answer. Uncertainty belongs in view; it should not be converted into either an invented defect or an automatic pass.
Attach reversibility to the decision without letting it replace severity. An easily reverted spacing adjustment may be cheap to defer or try later. A registration that silently creates duplicate records can leave effects that reverting the page will not remove. These are different consequences even if both repairs touch one line. The reviewer supplies the evidence about effects; the owner decides the permitted change and the release disposition.
Fix what prevents the task; classify the rest by evidence. AI-generated editorial illustration.
There is a legitimate objection to deferring polish: repeated small inconsistencies can make a product harder to understand. That deserves a coherent design task with its own acceptance conditions. It does not justify silently expanding every narrow bug fix into a redesign. If the current gap breaks an established requirement, cite that requirement. If it exposes a missing design rule, take the question to its owner before asking an assistant to enforce it.
For the next review, put the rubric beside the acceptance conditions and require a reason for every disposition. The hypothetical page comes back with submission blocked, focus requiring repair and the decorative gap deferred only after its lack of usability impact is established. A disagreement stays attached to the evidence and goes to the product owner. The review is finished when it supports that decision, not when the page has run out of things someone could prefer differently.
If this was useful, forward it to one engineer who needs less noise in their feed.


