Watching Real Sessions: A Plain Guide to Session Replay

A chart tells you that 40 percent of visitors left your pricing page without clicking anything. It does not tell you why. Maybe the page loaded with a cookie banner covering the button. Maybe the table was unreadable on a phone. Maybe the visitors found what they needed and left happy. The numbers cannot separate these cases.

Session replay can. It lets you watch a reconstruction of what one visitor did on your site: where the pointer went, what they clicked, how far they scrolled, which pages they moved through. This post explains what that means in practice, what it is good for, and where it misleads.

It is not a video

People often assume a replay is a screen recording. It is not. A recording would capture pixels from the visitor's screen, which is heavy and invasive.

Replay works differently. A small piece of script in the page records the structure of the page (the DOM) once, then records the changes that happen afterward: an element appears, text changes, a user scrolls, a pointer moves, a click lands. On your side, a player takes that stream of changes and rebuilds the page as it looked, replaying the changes in order.

Three consequences follow:

• Replays are small compared to video, so they are practical to collect across many sessions.
• Because the page is rebuilt, text and layout stay crisp, and sensitive fields can be masked before anything leaves the browser.
• The rebuild can be imperfect. Fonts, third-party widgets, canvas elements and cross-origin frames sometimes render differently from what the visitor saw. Treat a replay as a faithful sketch, not evidence in court.

What you actually see in OnyxMetric

In OnyxMetric, a replay shows pointer movement, clicks, scrolling, page navigation and form interactions with personal data masked. Next to the player you get the session details: country, device and browser. Replay is available on the Pro plan and you switch it on per site in the site settings. The tracking script picks the setting up by itself, so you do not edit your script tag to turn it on.

The point of keeping replay next to the rest of your analytics is context. You rarely start with "I want to watch a session." You start with something odd: a funnel step that dropped, a page with a high exit rate, an error that spiked. Then you ask for examples. Having sessions sit one click away from those reports is what makes replay useful rather than a time sink.

Three jobs where replay beats a chart

1. Finding the cause behind a drop. Say your signup funnel loses half its visitors between the form and the confirmation page. Open a handful of sessions that stopped there. You might see people tapping a disabled submit button because a password rule was never explained. No metric would have named that.

2. Reproducing a bug report. A customer writes "checkout doesn't work." Support asks for steps, the customer cannot remember them. If that session was recorded, you can watch the exact sequence, including the error state that appeared. Pair it with error tracking and you can go from a stack trace to the human path that triggered it.

3. Checking a design assumption. You moved a call to action below the fold because it looked cleaner. Watch ten sessions. If most people scroll past it at speed, or never reach it, you have your answer faster than an A/B test would give it.

Where replay misleads

Replay is persuasive, and that is the danger. A single dramatic session can steer a whole roadmap. Some habits keep you honest:

• Watch a set, not one. Look at five to ten sessions that share a trait before you decide there is a pattern. One rage-clicking visitor might be a bot or a person on a bad connection.
• Start from a question. "Why do mobile visitors abandon the cart?" gives you a filter and an end condition. "Let's see what people do" gives you an afternoon of scrolling.
• Do not confuse movement with attention. A pointer sitting still does not mean someone is reading. On touch devices there is no pointer at all.
• Count it afterward. If you see something in replays, check how common it is in the numbers. Replay generates hypotheses. Funnels, goals and event counts test them.

Privacy is a design constraint, not an afterthought

Recording what people do raises fair questions, and the answer should not be "trust us." Practical rules we recommend:

1. Keep PII protection on. OnyxMetric masks sensitive information in replays by default and offers extra masking of form inputs and text content in your settings.
2. Tell visitors. Your privacy policy should say that sessions may be recorded for product improvement, and describe what is masked.
3. Mask by default and unmask on purpose. Passwords, payment fields and anything medical or financial should never be visible. If a field is not sensitive and you need to see it, make that an explicit decision.
4. Limit who can watch. Replays are product research material. They do not belong in a shared channel or in a ticket attached to a person.
5. Do not use replay to identify individuals. If you need to understand a specific account, use identified user data deliberately, with the consent your region requires.

Cookieless counting and replay are different tools with different risk profiles. Counting pageviews without cookies is low risk. Recording interaction is a more intrusive activity, so treat the consent and disclosure question separately for it, and check the rules where your visitors live.

A small routine that works

You do not need a daily replay habit. A light weekly one does more:

1. Open your funnel and find the step with the biggest drop.
2. Filter to sessions that ended at that step.
3. Watch five of them with a notebook open. Write down what the visitor tried, what they expected, what happened.
4. Pick the one change that would have helped the most people and ship it.
5. Next week, check the same funnel step. If the rate did not move, watch five more.

Fifteen minutes, one change. Over a quarter that is a dozen fixes that came from watching real behavior rather than guessing at it.

When you can skip it

If your site is a few static pages and a contact form, replay is probably more than you need. Traffic sources, top pages and a goal on the form will tell you most of what you can act on. Replay earns its price when you have a flow with steps, a product people have to learn, or enough traffic that bugs hide in the noise.

If you do have those, turn replay on for a site, let it collect for a week, and start from your worst funnel step. You will see something you did not expect within the first handful of sessions. That is the point of watching.