Session Recording

Last updated 28 October 2026

Quick answer

mouse movement, scrolling, clicks, and the pages or states the visitor passed through. In an A/B test it shows how people behaved in a variant.

Key takeaways

A recording shows mouse, scroll, clicks, and page changes. It does not assign a variant or decide a winner.
Start with Microsoft Clarity. It is free, and Mida can tag converting sessions with the test and variant.
Hotjar is another recording option. Use it the same way: filter to the variant, then watch. Do not treat a handful of replays as the result.

Definition

mouse movement, scrolling, clicks, and the pages or states the visitor passed through. In an A/B test it shows how people behaved in a variant. It does not assign the variant, and it does not pick the winner.

What Session Recording means in A/B testing

The recorder saves a timeline of the DOM and the pointer, then plays it back. Heatmaps are the aggregate of many of those sessions. A recording answers "what did this person do?" A metric answers "how often did the outcome happen?" You need the second to call a test. You use the first when the metric moved and you cannot see why from the rate alone. Mida does not record sessions itself. It connects the experiment to a recorder by tagging the session.

Why Session Recording matters

A losing variant can be a bad idea or a broken control. Ten replays of the losers often show a covered button or a field that never accepts input. A winning variant can be a lift you do not understand, which makes the next test a guess. Recordings are how you pick that next test. They are a poor place to discover the result, because the sessions you happen to open are not a random sample of the metric.

Example of Session Recording

Add-to-cart is down on variant B. In Clarity you filter to the Mida conversion tag for that variant and watch the sessions that did add to cart, plus, if you are sampling the page generally, how far people scroll before they leave. The new size chart opens over the button on mobile. The rate told you the variant lost. The replays tell you the chart is covering the control. You fix the overlap and rerun. You do not ship the old chart because three replays looked painful.

How to use Session Recording

Install Microsoft Clarity and connect it to the experiment. Clarity is the one to use first because it is free and does not impose a recording quota. Confirm a test conversion writes the variant tag, then filter replays and heatmaps to that tag after the test has enough conversions to be worth watching. Hotjar is an alternative if your team already reviews sessions there. Mida can tag converting Hotjar sessions too. Either way, write down what you saw and turn it into the next hypothesis. Do not average a dozen replays into a win rate.

Common mistake

A common mistake is watching unfiltered recordings of the whole site and calling that the variant research. Without the test and variant on the session, you are mixing arms. Another is storing recordings of fields that contain personal data you did not mean to capture. Mask inputs that are not required to see the layout problem.

How Mida works with session recording

Mida does not include a session recorder. The documented path is to tag the recorder you already run. For Clarity, which is the free option and the one to set up, the Clarity integration calls clarity('set', 'mida_conversion', 'mida_conversion_{test_id}_{variant}') when a visitor converts. You then filter Clarity's recordings and heatmaps by that tag. Details and the reason exposures are not tagged are on the Microsoft Clarity glossary page.

Hotjar is the other documented option. Its integration sends conversion event names rather than a Clarity custom tag. Use it when Hotjar is already where the team watches sessions. It is not required if Clarity is in place.

Related A/B testing terms

FAQ

Which recording tool should I connect first?

Microsoft Clarity. It is free, it does not cap recordings, and the Mida integration can mark a converting session with the test and variant. That is enough to replay the people who converted on each arm. Add another recorder only when you already rely on it for a workflow Clarity is not covering.

Does a recording replace the conversion count?

No. The count, on the metric you pre-registered, is the result. The recording is how a visitor moved through the variant. Watch recordings after you know which arm is ahead or behind, so you are looking for a mechanism rather than hunting for a winner in a few sessions.

Can I record every visitor who was assigned to variant B?

Not with the standard Mida tags. Both the Clarity and Hotjar integrations mark the session when the visitor converts, not merely when they are exposed. That keeps the filter pointed at people who reached the goal. Exposure-level recording is not part of those integrations.