QoE: measure what the viewer experiences
Learn to: define viewing metrics and diagnose a symptom at its responsible layer.
Prerequisites: 15-lifecycle-accessibility, 16-security-delivery
Failure ownership map
On a narrow screen, swipe the diagram horizontally to keep its labels readable.
Metric definitions and cohorts
On a narrow screen, swipe the diagram horizontally to keep its labels readable.
Mental model synthesis
Quality of experience combines startup, stalls, errors and the delivered viewing quality. Mux Data documents those dimensions. An app must state metric definitions before comparing cohorts; a network status code alone does not describe the viewing experience.
[S35]Worked diagnostic synthesis
A lesson starts on desktop but not mobile Safari. Compare media type/codec, native versus MSE path, autoplay rejection and license support. If the first frame appears then stalls, inspect segment timing, buffered duration and selected bitrate. Keep UI errors, engine errors and delivery errors distinct.
[S03] [S04] [S16] [S23] [S27]Proposed measurement contract synthesis
Record asset ID, player/build version, browser/OS, delivery path, ready/play/playing/waiting/error milestones and coarse session correlation. Define attempted playback explicitly. Decide whether startup and user pauses belong in the rebuffer denominator. These are proposed app semantics, not copied vendor formulas.
[S03] [S35]Decision and improvement synthesis
Use a managed QoE integration when its coverage and privacy controls fit; otherwise own a small event contract and browser probes. Sample noisy events and avoid bearer URLs or personal content in logs. Test one suspected layer at a time; retries should not reset every metric and make a failure look like a new success.
[S16] [S35]Check yourself: Why report startup p95 by browser rather than only an overall average?
Averages can hide a device-specific problem. Cohort distributions show which users encounter long startup and help locate the responsible path.