产品 UX/UI 体验问题审计
沿关键用户旅程识别界面与交互中的阻碍,区分可证实的可用性问题和主观视觉偏好,并整理改进建议。
完整提示词
I want you to perform a maximally deep, systematic, evidence-first and product-oriented audit of the complete UX/UI experience of the application or digital product. Main objective: > Determine where the interface, product structure, navigation, feedback, state handling or visual hierarchy make it harder for the user to understand the system and successfully complete their real goal, while clearly distinguishing confirmed usability problems from subjective design preferences. This is not: - a generic "make it cleaner" - redesign for the sake of redesign - criticizing colors without context - insisting on a particular design trend - automatically declaring every extra click a problem - automatically recommending minimalism - automatically recommending more elements - a replacement for an accessibility audit - a replacement for product strategy Priority: **task failure > destructive/confusing action > inaccessible critical flow > lost user state > navigation failure > unclear system status > preventable user error > cognitive friction > consistency > aesthetics** It is better to find 10 UX problems that actually block the user than 100 visual preferences. ## 1. PRODUCT CONTEXT Before any findings, establish: - what the product does - who uses it - the primary user personas - the most important tasks - business goals - monetization, if relevant - device context - frequency of use - novice vs expert usage ## 2. CRITICAL USER JOURNEYS Map the real flows. For each: ```text Journey: Actor: Entry point: Goal: Required information: Actions: Decision points: Persistent state: Exit/success: Failure/recovery: ``` ## 3. FIRST IMPRESSION Does the user understand: - what the product is - what they can do - where to start - what the next step is ## 4. INFORMATION SCENT A link or button label should predict its destination well enough. ## 5. VISUAL HIERARCHY Does attention go to: - the main task - the most important status - the primary action ## 6. PRIMARY ACTION Clear. ## 7. SECONDARY ACTION It must not visually outweigh the critical primary action without a reason. ## 8. DESTRUCTIVE ACTION Distinguish it. ## 9. AFFORDANCE Does an element look interactive when it is? ## 10. FALSE AFFORDANCE Does a decorative element look clickable? ## 11. LABELS Terminology should match the user's mental model. ## 12. INTERNAL JARGON Do not expose it unnecessarily. ## 13. CONSISTENCY The same concept should behave the same way. ## 14. EXCEPTION Consistency is not the goal if different behavior has a clear reason. ## 15. USER CONTROL Back/cancel/undo where relevant. ## 16. SYSTEM STATUS The user must know: - loading - saving - success - failure - pending - offline - sync ## 17. LATENCY UX Prevent uncertainty. ## 18. SKELETON Not automatically better. ## 19. SPINNER Not automatically bad. ## 20. OPTIMISTIC UI Audit correctness + rollback. ## 21. PESSIMISTIC UI Potential unnecessary waiting. ## 22. EMPTY STATE Should answer: - why empty - what next ## 23. ZERO DATA ## 24. LOADING STATE ## 25. ERROR STATE ## 26. PARTIAL ERROR ## 27. RETRY ## 28. UNKNOWN OUTCOME Critical. ## 29. OFFLINE ## 30. SYNC ## 31. STALE DATA ## 32. CONFLICT ## 33. FORM Check labels, validation timing, error recovery, preservation of entered data and the submit outcome (double submit, unknown outcome). ## 34. NAVIGATION Check whether the user knows where they are, whether labels predict the destination and whether back, URLs and deep links keep the expected state. ## 35. MOBILE UX Check touch targets, the keyboard overlay, interruptions (background/resume), network loss and platform conventions for back. ## 36. ACCESSIBILITY Check whether critical flows can be completed with a keyboard only and with a screen reader, with visible focus and associated error messages. A detailed WCAG audit is outside the scope of this prompt. ## 37. ONBOARDING Check whether a new user reaches the first real value without premature setup and decisions they do not yet understand. ## 38. CRITICAL FLOW For each critical flow, follow every step from the entry point to the business outcome, including interruption, retry and recovery. ## 39. ERROR PREVENTION ## 40. ERROR RECOVERY ## 41. VALIDATION ## 42. CLEAR ERROR MESSAGE Should tell user: - what happened - where - how to fix ## 43. FOCUS After error, focus should help recovery where appropriate. ## 44. CONFIRMATION Not every action needs modal. ## 45. HIGH-RISK ACTION Confirmation may be justified. ## 46. UNDO Often better than blocking confirmation when reversible. ## 47. DATA LOSS Protect: - unsaved form - draft - navigation away ## 48. MULTI-STEP FLOW Progress. ## 49. WIZARD Not automatically preferable. ## 50. OPTIONAL STEP ## 51. BRANCHING FLOW ## 52. SKIP ## 53. RETURN LATER ## 54. PERSISTENCE ## 55. USER MEMORY BURDEN Avoid forcing user to remember data visible elsewhere. ## 56. RECOGNITION VS RECALL ## 57. INFORMATION DENSITY Depends on user/task. ## 58. POWER USER Dense UI can be correct. ## 59. NOVICE USER May need guidance. ## 60. PROGRESSIVE DISCLOSURE ## 61. SHORTCUT ## 62. BULK ACTION ## 63. MULTISELECT ## 64. FILTER ## 65. SEARCH ## 66. SORT ## 67. PAGINATION ## 68. INFINITE SCROLL Neither automatically better. ## 69. TABLE ## 70. CARD Use data/task evidence. ## 71. RESPONSIVE ## 72. TOUCH TARGET ## 73. HOVER-ONLY ## 74. KEYBOARD ## 75. SCREEN SIZE ## 76. ZOOM ## 77. CONTENT LENGTH ## 78. TRANSLATION Longer localized strings. ## 79. DATE/TIME ## 80. CURRENCY ## 81. NUMBER ## 82. USER TRUST Critical when: - payments - privacy - destructive actions - security ## 83. PRIVACY UX Do users understand what is shared? ## 84. PERMISSION PROMPT Ask at meaningful time. ## 85. AUTH Session expiry should preserve intent where safe. ## 86. PAYWALL Should be clear before work is lost. ## 87. UPGRADE ## 88. LIMIT User should know quota state. ## 89. DISABLED CONTROL If disabled, ideally user understands why. ## 90. TOOLTIP Not replacement for essential content. ## 91. MODAL Not automatically bad. ## 92. TOAST Not suitable for every critical message. ## 93. NOTIFICATION ## 94. PERSISTENT STATUS ## 95. ACCESSIBILITY OF FEEDBACK ## 96. USER TEST EVIDENCE If available, prioritize: - usability study - support tickets - analytics - session recordings - funnel drop-off over personal preference. ## 97. ANALYTICS Can show where issue occurs, not always why. ## 98. DROP-OFF Investigate cause. ## 99. RAGE CLICK Signal, not proof. ## 100. SUPPORT TICKET Strong contextual evidence. ## 101. HEURISTICS Can use established heuristics, but not as mechanical scorecard. ## 102. FALSE POSITIVE RULES Do not automatically report: - modal - dropdown - hamburger menu - dense table - long form - multiple clicks - disabled button - confirmation dialog A problem must have a concrete user/task consequence. ## 103. EVIDENCE TIERS ```text A - observed user failure, usability test, analytics/support evidence B - complete user-flow evidence demonstrating friction/failure C - strong heuristic/accessibility evidence D - plausible usability concern requiring validation E - stylistic/hardening suggestion ``` ## 104. STATUS ```text CONFIRMED LIKELY NOT VERIFIED CONTROLLED NOT APPLICABLE HARDENING ``` ## 105. SEVERITY P0: - UX causes catastrophic irreversible loss or critical unsafe action at scale P1: - critical user task frequently fails or destructive action is easily triggered/misunderstood P2: - material completion/friction/accessibility problem P3: - limited usability inconsistency P4: - polish/hardening ## 106. FINDING FORMAT ```text ID: Severity: Status: Evidence tier: Persona: Journey: Screen/state: User goal: Current behavior: Expected mental model: Friction/failure: Impact: Evidence: Root cause: Recommended change: Alternative: Validation method: Regression risk: ``` ## 107. MATRICES ### Journey Matrix | Journey | Entry | Main action | Error recovery | Completion | |---|---|---|---|---| ### State Matrix | Screen | Loading | Empty | Error | Success | Offline | |---|---|---|---|---|---| ### Interaction Consistency Matrix | Concept | Screen A | Screen B | Difference justified | |---|---|---|---| ## 108. SECOND PASS Repeat the audit as: - first-time user - expert user - keyboard-only user - mobile user - user with slow network - user returning after session expiry - user with empty account - user with large dataset - user who makes a mistake halfway - user who cancels/retries ## 109. FINAL QUALITY GATE Confirm: - product context - personas - critical flows - hierarchy - navigation - feedback - loading - error - empty - destructive actions - data loss - forms - mobile - accessibility - trust - localization - recovery - evidence ## 110. OUTPUT `ULTIMATE_UX_UI_PRODUCT_AUDIT.md` ## 111. FAILURE CHAINS ```text user clicks Delete ↓ confirmation dialog says only "Are you sure?" ↓ does not identify selected workspace ↓ user has two similarly named workspaces ↓ confirms wrong target ↓ irreversible data loss ``` ```text user submits long form ↓ spinner appears ↓ request times out after server actually saved data ↓ UI shows generic error ↓ user retries ↓ duplicate entity created ``` # FINAL RULE A UX finding is not: > "I would have designed this differently." A UX finding must show: ```text user goal + current interaction + specific friction/failure + measurable or logically justified consequence ```
分类与标签
设计创意、UX审计、UI体验、用户旅程、产品可用性