I’m trying to make a practical checklist for finishing a UI. I’ve been looking at tap targets, animation performance, the number of choices on screen, and testing light and dark themes at mobile and desktop sizes. What would you add, and which rules are too rigid?
What checks tell you a UI is actually done?
Answers
Start with the paths people actually need to complete. Can you reach and operate every control with a keyboard, and can you see where focus is? Check text contrast with a tool rather than trusting your eyes.
Then zoom the browser to 200% and try larger system text. Check some widths between phone and desktop too. A layout can look fine at both ends and fall apart in the middle.
Reply to this answer
Do a break it pass. Try long names and titles, no results, lots of results, a slow connection, and a failed request. Loading, empty, error, offline, and partial failure states should make sense instead of leaving someone staring at a spinner.
Use real content where you can. Placeholder text is often much kinder to a layout than actual names, prices, or translations.
Reply to this answer
I wouldn’t make four choices a hard limit. A dashboard may need filters, bulk actions, and table settings all at once. Check instead whether the main action is clear and whether the other controls make sense for the task.
Removing clutter is useful, but not if you remove something people need to get the job done.
Reply to this answer
Be careful with the units in the tap target rule. Apple’s 44 by 44 guidance is in points, while Android’s 48 by 48 guidance is in density-independent pixels. Those aren’t automatically the same as CSS pixels.
Likewise, limiting animation to transform and opacity is a useful implementation check, not something you can confirm just by looking at the finished screen. Check how animation performs on the devices you support rather than treating the property rule as proof of a steady frame rate.
Reply to this answer
Check on actual phones if you have access to them, not just a resized browser window. Mobile browsers can behave differently. A simulator can help catch issues too, but it’s another check rather than a complete substitute for a device.
Zelvexa.3n:
A slower phone and a throttled connection are useful additions to that pass, especially for loading behavior.
Reply to this answer
Add your answer
Answers are reviewed before publication.
Rovixa-8j:
For contrast, WCAG AA gives you a concrete target to check against. Make sure the focus indicator stays visible as well.