Unmoderated Usability Testing With Session Replay

Short answer: give five people a task and a link, let them try it on their own device, and watch the recordings afterwards. A tagged link lets you find their sessions. This costs almost nothing and shows where a new person gets stuck. It will not capture what they were thinking, so add a couple of follow-up questions. Use it alongside recordings of real traffic, which show what happens when nobody is being asked to try.

Table of Contents

1. Three Ways to Watch Users
2. Setting Up a Test
3. Writing Good Tasks
4. Analysing the Sessions
5. Limits
6. Key Takeaways

Three Ways to Watch Users

Unmoderated testing sits in between. It is especially useful before launch, when there is no real traffic yet, and for flows real visitors rarely reach.

Setting Up a Test

1. Pick one flow. Signup, finding a product, booking, a new feature.
2. Recruit five people. Customers, friends of the right profile, or people from your waiting list. For a first pass, anyone who has not seen the product.
3. Create a tagged link so their sessions can be found:

https://example.com/?utm_source=usability-test&utm_campaign=signup-flow-oct

1. Send a short brief: the task, the link, and a request to use their normal device.
2. Tell them it is recorded. Say that their screen activity on your site is captured, that typed text is masked, and get their agreement. Make sure your consent setup allows recording for them.
3. Do not explain the product. The point is to see what happens without help.
4. Ask two or three questions afterwards by message or a short form.

Then filter sessions by the UTM source and watch.

If you want to keep test sessions out of your normal reports, use a staging site with its own project, or simply exclude the campaign when analysing.

Writing Good Tasks

A task should describe a goal, not the steps.

Rules:

• Use their words, not your interface labels. If the button says "Workspaces", do not use that word in the task.
• One task at a time. Two or three in a session at most.
• Give a realistic scenario.
• Define success for yourself beforehand.
• Do not hint.

Follow-up questions:

1. What, if anything, was confusing?
2. Was there a point where you nearly gave up?
3. What did you expect to happen that did not?

Analysing the Sessions

For each participant, note:

Then compare across participants. A problem seen in three of five is very likely real. One seen once may be personal.

DeepSync's Highlights and session signals help find the moments quickly: rage clicks, errors, and dead ends are surfaced in the player (session replay).

Combine the recording with the participant's answers. "I looked for pricing in the menu" plus a recording of them opening the menu and closing it is a complete finding (feedback and behavior).

Five participants is a widely used starting point for finding the most common problems in one flow. Test, fix, and test again with five more, instead of testing twenty at once.

Limits

• No thinking aloud. You see what they did, not why. Add questions, or run a few moderated sessions for depth.
• Artificial motivation. A test participant persists longer than a real visitor would. Real traffic gives up sooner.
• Recruits are not your market unless you choose them carefully.
• Small numbers. Good for finding problems, useless for measuring rates.
• Capture limits. Content in cross-origin iframes and on canvas is not recorded (capture limits).
• Consent. Participants must agree, and personal data must be masked as for any visitor.

Use unmoderated tests to find problems, and real-traffic recordings to learn how often those problems occur.

Key Takeaways

• Five people, one task, a tagged link, and their recordings.
• Describe a goal in the participant's words, not the steps.
• Compare across participants and add follow-up questions.
• Use tests to find problems and real traffic to size them.

Conclusion

Usability testing does not need a lab or a budget. It needs five people who have never seen your product and the willingness to watch them struggle.

Start free with DeepSync.