UX Principles & Heuristics
The evaluation backbone. Every design review should walk these.
Nielsen's 10 heuristics (operationalized)
- Visibility of system status — every action gets feedback within 100ms; loading >1s shows progress; background work shows state (saving…/saved).
- Match to the real world — user vocabulary, not system vocabulary; natural ordering; familiar metaphors.
- User control & freedom — undo > confirm; back always works; easy exit from every flow; cancel long operations.
- Consistency & standards — internal consistency (one pattern per job) and platform conventions (don't reinvent share/back/tabs).
- Error prevention — constraints over corrections: good defaults, disabled-invalid combinations explained, confirm irreversible, autosave.
- Recognition over recall — show options, don't make users remember codes/paths; recently-used surfaces; visible state.
- Flexibility & efficiency — shortcuts for experts (keyboard, gestures, command palette) that stay invisible to novices; smart defaults for everyone.
- Aesthetic & minimalist design — every element competes for attention; cut what doesn't serve the current task.
- Help users recognize & recover from errors — plain-language errors: what happened, why, what to do; preserve user input.
- Help & documentation — contextual help at point of need (tooltips, inline hints) beats manuals.
Behavioral laws to apply
- Fitts's Law: important targets = big and close. Thumb zone on mobile (bottom half); screen edges/corners on desktop are "infinite" targets.
- Hick's Law: decision time grows with options. Progressive disclosure; recommended defaults; 3-tier pricing not 7.
- Jakob's Law: users spend most time on other sites — meet conventions, spend novelty budget only on your core differentiator.
- Miller's Law / chunking: group into 5–7 chunks (phone numbers, card numbers, nav sections).
- Peak–end rule: invest in the flow's peak moment and its ending (success states, confirmations, celebrations).
- Von Restorff: the one visually-different element gets remembered — that's why one primary CTA works.
- Doherty threshold: keep system response <400ms to keep users in flow; use optimistic UI where safe.
- Goal-gradient: show progress; users accelerate near completion (progress bars start pre-filled ~10%).
- Tesler's Law: complexity is conserved — absorb it in the system (autodetect, defaults) rather than pushing to users.
Cognitive load rules
- One primary job per screen. Screens that do everything do nothing.
- Progressive disclosure: advanced options behind "Show more"; wizard steps over mega-forms.
- Defaults do the work: 90% of users never change them — make defaults the recommended path.
- Blank-slate anxiety: never open on an empty canvas without a starting template/example.
Feedback timing budget
| Delay | Requirement |
|---|---|
| <100ms | Feels instant; no indicator needed |
| 100ms–1s | Subtle feedback (button state), no spinner |
| 1–3s | Spinner/skeleton |
| >3s | Progress + explanation; allow cancel |
| >10s | Progress + time estimate; allow backgrounding + notify |
Skeletons beat spinners for content layout; optimistic UI (assume success, reconcile on failure) beats both for social actions.