What is session replay?
Session replay is a technique that records a user's interactions and DOM state so their session can be replayed as video. It is useful for debugging specific reported issues, and it carries real trade-offs in scale, analysis cost, and privacy.
Session replay records what happens in a user's browser, the DOM, the clicks, the scrolling, the typing, and reconstructs it so someone on your team can watch the session back like a video. Vendors capture DOM mutations rather than literal screen video, but the product is the same: a recording of a person using your site.
We should say up front that Flusterduck does not do session replay. This page explains the technique fairly anyway, because the trade-offs are real on both sides.
What replay is good at
Replay earns its keep in two situations.
Debugging a specific report: a customer says checkout is broken on Safari, you filter recordings to Safari plus checkout plus yesterday, and you watch exactly what they saw. That is targeted investigation with a known query, and replay is excellent at it.
Building empathy: engineers watching five real sessions during onboarding often change how they think about the interface. That is a calibration exercise, and a valuable one.
Where replay struggles
The trouble starts when replay becomes the primary discovery tool, because discovery means watching. If a fraction of one percent of sessions contain a friction moment worth fixing, finding ten of them means sifting thousands of recordings. Nobody does that, so in practice teams watch a handful and generalize, which means fixing whatever the last watched recording happened to show. Recordings do not aggregate: you cannot ask a pile of videos which element frustrated the most people this week.
The privacy surface is the other cost. A recording captures whatever was on screen, and masking is brittle: a new field, a third-party iframe, a dynamic element, and the recording contains something it should not. Regulators treat screen recordings as personal data collection, consent banners cut the sample, and stored recordings are a liability that compliance teams increasingly ask about. The longer argument is in Session replay won't tell you what to fix.
The alternative: record the behavior, not the screen
The struggle replay hopes a human will notice can be detected directly. Frustration signals, rage clicks, dead clicks, thrashing, loops, and field-level abandonment, aggregate cleanly, attach to specific elements, rank by how many people they touched, and carry no video, no keystrokes, and no personal data. You lose the movie. You gain a ranked, located list of what to fix and a way to verify the fix worked.
Teams that keep replay for targeted debugging and use signals for discovery get the best of both. Teams choosing one tool for finding what to fix should know the watching does not scale. A free scan shows what signal-based detection finds on your site.