Tooling · 3 Aug 2026
Session replay blind spots your tooling never flags
Replay libraries feel omniscient. They are not. Treat them as partial witnesses.
Teams in our cohorts often arrive convinced their stack “records everything.” Then a privacy mask hides the exact field that stalled a shopper, or sampling quietly drops slow networks common on weekend mobile traffic.
Sampling that favors the easy sessions
If capture throttles under CPU pressure, you under-observe the very devices that struggle. Cross-check replay counts against server-side page views before you celebrate a friction “win.”
Masks that erase the plot
Masking payment fields is necessary. Masking labels beside them is optional — and often accidental. When labels disappear, hesitation looks like random mouse fidgeting.
SPA route lies
Client-side routers can report a single “page” while users traverse three mental spaces. Anchor your friction tickets to route changes your engineers recognize, not just tool-generated titles.
Practice
Once a month, watch five sessions from the slowest quintile of connection speed. The narrative differs from the median samples your dashboard surfaces first. That habit ships inside our Repair Backlog Orchestra track.