Usability Testing — Watch Real Users, Not Test Scripts

Usability bugs are invisible to normal testing. The button works — it just isn't where users expect it. The flow completes — it just takes 3x longer than it should.

None of these show up in functional tests. None appear in automated regression. They only surface when a real human tries to use your product for a real task.

What usability testing actually tests

Findability — Can users find the feature?

Clarity — Do they understand what to do next?

Efficiency — How many steps does it take?

Error recovery — What happens when they make a mistake?

Trust — Do they feel confident completing the task?

Moderated vs unmoderated

Moderated: researcher watches + asks questions. Highest quality.

Unmoderated: user completes tasks alone, screen + voice recorded. Scalable.

How to run a session

Define the task — "sign up and connect payment" not "test signup"

• Recruit real users matching your base
• Give them the task, not instructions
• Stay silent — the struggle is the data
• Record everything (screen, voice, expression)
• Ask non-leading questions after

What we typically find

• Buttons that don't look clickable
• Forms with unnecessary fields
• Confusing error messages
• Missing confirmation feedback
• Internal naming used in UI

Metrics to track

• Task success rate
• Time on task
• Error rate
• Satisfaction score
• Think-aloud friction

When to run usability testing

• Before major changes
• After new features ship
• When metrics decline without obvious cause
• During redesigns
• Quarterly

Common mistakes

• Testing colleagues instead of real users
• Leading users toward the "right" answer
• Skipping recruitment screening
• Not recording sessions
• Ignoring findings