Session replay is genuinely good at one thing: showing you what happened when nobody can describe it. If you are debugging a conversion drop with no reports and no errors, watching sessions is sometimes the only way in.

That is a narrower job than the way it is usually bought. A lot of teams end up with a recorder on every page because it came bundled with the feedback widget they actually wanted.

What the bundle costs

Turning on a recorder is not a neutral act. It brings:

A consent obligation in most jurisdictions, because you are storing behavioural data about identifiable sessions
A masking configuration you have to maintain, so recordings do not capture card numbers, health information, or private messages
A substantially larger script that runs continuously rather than on demand
A new store of sensitive data you now have to secure, retain, and delete on request
A conversation with your security reviewer that you did not previously have to have

The honest counterargument

Without a recording, some bugs get harder. A report that says "it broke" and nothing else is a dead end you could have watched your way out of.

The mitigation is not free either — it is a form that asks better questions, an optional email so you can go back to the reporter, and a screenshot the reporter can attach on a press. That works for most reports. It does not work for all of them, and pretending otherwise would be dishonest.

The screenshot deserves the same scrutiny as everything above, because it borrows one of the costs on that list: if a page can render itself to an image, somebody has to decide what must not appear in it. The difference in scale is the argument. A recorder needs a masking policy that holds for every second of every session nobody is watching; a single frame taken on an explicit press, blanked before it is drawn and shown to the reporter before it is sent, needs one that holds for a moment they chose. Smaller, but not nothing — and a tool that claimed otherwise would be doing the thing this post is complaining about.

The actual decision

Ask what fraction of your reports genuinely required watching a replay to resolve. For most teams the number is small enough that the recorder is being carried by a handful of cases a year.

If that is you, the feedback button on its own is the better trade: no consent banner, no recordings to protect, and a script small enough that nobody has to defend it in a performance review.

And if you do need replay — keep it. These are separable decisions, which is the whole point.