A/B Testing

Test two or more iterations or versions of a design which enables designers to test out different ideas with users to choose a better option.

Why it matters

It settles arguments with evidence instead of seniority: the design that wins is the one that performed, not the one whose author spoke loudest. Its trap is the local maximum — endless button-colour tests can polish a design that should have been rethought.

The numbers

  • 2+ variants, one variable — change more and you cannot say what won
  • Significance first — call the winner too early and noise picks it

Related terms

Go deeper in Learn UX

Ask your assistant

Define "A/B Testing" using uxspot.

Works in Claude, Cursor or any MCP client — connect the uxspot MCP.

Think it stuck? A 20-second round of Spot Check will tell you.