dembrandt Status Colours and Error Design
Keep the Colour Set Small
Every status colour added to a system is a cognitive burden on the user. They must learn what each colour means, remember it, and interpret it correctly under stress — which is exactly when errors occur.
The minimal set that covers almost everything:
| Colour | Semantic meaning | Always means |
|---|---|---|
| Red | Error / failure / destructive | Something went wrong, or this action cannot be undone |
| Orange / Amber | Warning | Something needs attention before proceeding |
| Green | Success / positive | Action completed, state is healthy |
| Blue | Info / neutral status | Informational, no action required |
Rule: each colour maps to exactly one meaning across the entire product. If orange means "warning" in one component and "pending" in another, the system breaks down.
When in doubt, cut the colour — neutral grey communicates status without semantic weight, and neutral is better than a misused semantic colour.
Orange Is Always a Warning
Orange (amber) carries a specific signal: pay attention, something may go wrong. Do not use it for:
- Neutral states (use grey)
- Progress or pending (use blue or a spinner)
- Informational content (use blue)
- Branding or decorative purposes inside status indicators
If orange appears in the UI, the user should immediately know it requires their attention.
Errors Should Be Recoverable
The worst error is one the user cannot recover from. Design every error state with a path forward.
Error message anatomy
Every error should answer three questions:
- What went wrong? — plain language, no error codes
- Why did it happen? — if useful and known
- What should the user do next? — specific, actionable
❌ "Error 500"
❌ "Something went wrong"
✓ "We couldn't save your changes. Check your connection and try again."
✓ "This email is already in use. Sign in instead, or use a different email."
Recovery actions
- Always provide a retry button for transient failures (network errors, timeouts)
- For validation errors, point directly to the problematic field
- For destructive actions that failed, reassure the user nothing was lost
- For session expiry, redirect to login and return the user to where they were
Prevent Large Errors Before They Happen
The most damaging errors — data loss, irreversible actions, broken state — should be prevented at the design level, not handled after the fact.
- Confirm before irreversible actions: "Delete this project and all 47 tasks? This cannot be undone."
- Disable unavailable actions rather than letting users trigger them and hit an error
- Autosave where possible so a browser crash or accidental close does not destroy work
- Optimistic UI with rollback: show the success state immediately, silently retry on failure, surface an error only if the retry also fails
State the consequences before the user acts, not after. Fear comes from not knowing what a control will do. Make the outcome legible ahead of the action so the user commits with confidence:
- Plain-language consequence text next to or inside the control: "This will email all 240 subscribers", "Publishing makes this visible to everyone".
- Staged guidance for anything multi-step or heavy: break it into steps and tell the user what each one will do before they proceed, so the whole operation is predictable rather than a leap.
- Reserve the strong interruptions (confirm dialogs, typed confirmation) for the genuinely dangerous cases above; for everyday actions, a quiet line of helper text is enough. The goal is the same — no surprises, so nothing feels scary.
Levels of Severity — Use Sparingly
Not every problem is equal. Match the visual weight of the feedback to the severity.
| Severity | Pattern | When to use |
|---|---|---|
| Blocking error | Full-page error state or modal | App cannot continue, user must act |
| Inline error | Red text below a field | Form field validation failure |
| Toast / snackbar | Temporary notification, bottom of screen | Transient failure the user should know about but can dismiss |
| Alert banner | Persistent bar at top of section | Ongoing issue affecting the current context |
| Empty state | Illustrated or descriptive empty screen | No data yet — use as an opportunity for guidance, not just a blank |
Avoid showing multiple simultaneous error types at once — one clear message is more useful than three overlapping alerts.
Review Checklist
- Does the product use four or fewer semantic status colours?
- Does each colour mean exactly one thing, used consistently everywhere?
- Is orange/amber reserved exclusively for warnings?
- Does every error message state what went wrong and what to do next?
- Do all transient errors (network, timeout) have a retry action?
- Are irreversible destructive actions protected by a confirmation step?
- Is autosave or draft recovery available for long-form or complex inputs?
- Are multiple simultaneous error states avoided?
Common Anti-Patterns
| Anti-pattern | Problem | Fix |
|---|---|---|
| "Something went wrong" with no action | User is stuck with no path forward | Add specific cause and a retry or contact link |
| Orange used for "pending" and "warning" simultaneously | Colour loses its meaning | Orange = warning only; use blue or spinner for pending |
| Five or more status colours (red, orange, yellow, teal, purple…) | User must learn and remember a complex legend | Cut to the minimum: red, orange, green, blue |
| Inline validation only on submit | User discovers all errors at once | Validate on blur (leaving a field) for immediate feedback |
| No confirmation on delete | Users accidentally delete data | Require explicit confirmation for all irreversible actions |