Myths: Unsupported and overgeneralized website design rules
Purpose: stop memorable shortcuts from replacing their evidence. Each myth gets one of three verdicts: contradicted (evidence or the standard says otherwise), unsupported (no evidence for it as a general rule), or conditional (true only in a stated context). A short qualifier after the verdict says how far it goes; keep it when you cite the verdict. Verdict labels set 29 October 2026; the three-click entry was rechecked at its source that day, and the others rest on the reading of 13 September 2026.
A task must take no more than three clicks. Verdict: contradicted as a universal limit. NN/g reports that no published study supports the rule and that a study by Joshua Porter found no rise in dropoff after three clicks (The 3-click rule for navigation is false, read 2026-10-29). Check understandable routes and successful finding rather than enforce an arbitrary depth. Basis: NN/g's navigation review.
A menu must have at most seven items. Verdict: unsupported; the memory slogan does not establish it. Do not infer a navigation limit from a remembered chunk-capacity number. The original memory experiments have not been reviewed in this organization. Assess grouping, labels, and the task instead. Basis: navigation guidance and the scope of Shneiderman's memory advice.
Five users prove the design is usable. Verdict: conditional; overgeneralized as stated. Small iterative qualitative rounds are not exhaustive and do not establish a conversion uplift. Distinct users and tasks need coverage; quantitative precision requires its own plan. Basis: qualitative scope and quantitative scope.
WCAG 2.2 AA requires every target to be 44 by 44 CSS pixels. Verdict: contradicted; incorrect for AA. AA SC 2.5.8 uses 24 by 24 CSS pixels with specified exceptions. The 44 by 44 criterion is enhanced AAA, also with exceptions. Larger targets may still be desirable; do not change the conformance claim to justify them. Basis: AA target size and AAA target size.
Every interaction is fine if it finishes within 400 milliseconds. Verdict: unsupported; no universal specification established here. Input acknowledgment, browser rendering, and completed remote work are different measurements. Tognazzini's 50-ms advice is historical expert guidance, not another competing normative deadline. Basis: responsiveness versus completion and historical acknowledgment guidance.
A prettier screenshot proves better usability or conversion. Verdict: unsupported; an invalid transfer of evidence. Aesthetic ratings support conclusions about aesthetic judgment under the study conditions, not a product's downstream outcomes. Test impression and task use separately. Basis: limits of screenshot evidence.
Simpler always means hiding controls or placing one question on each page. Verdict: conditional; it depends on the context. Related choices may need simultaneous visibility; secondary functions need a discoverable path. GOV.UK explicitly allows related questions together when research supports that choice. Basis: disclosure distinctions and question-page qualifications.
A reassuring status, undo button, or AI explanation proves control. Verdict: unsupported; the interface alone does not establish it. Our rule is to verify the actual effect, available recovery, and evidence. Do not claim an external effect is reversible or an output is correct solely because the interface implies it. This is a design and implementation boundary, not a separate controlled finding. Basis: conditional reversal and Human-AI limits.
Famous historical revenue stories provide our expected uplift. Verdict: unsupported; unverified and not transferable without local evidence. Revenue anecdotes that circulate in design writing are not planning inputs here. No claim is made that they are false; their primary records and transfer conditions have not been verified in this organization. Use an explicit hypothesis and local measurements. Method basis: goal-based measurement.