Choose a reproducible interval
For a viewer comparison, start timing at the explicit Play action and stop at the first observed moving programme picture. Record sound separately, because audio and video can begin at different moments. Keep any advertisement or pre-roll distinction clear if it changes what first programme picture means.
Imagine a poster visible at 0.2 seconds, sound at 1.5 seconds and the first moving picture at 2.0 seconds. Reporting 0.2 seconds as video startup would hide most of the actual delay. The example values are illustrative and do not represent a measurement of this service.
Do not mix different test states
A returning session may reuse cached resources or an existing connection, while a fresh session may need more setup. Note which state applies. A comparison between a warm repeat and a fresh start can still be useful, but the labels should make that difference visible.
Keep device, browser, representation and observation conditions comparable. Several runs can show variability, but avoid publishing a single fastest result as a universal promise. If the application adapts quality during startup, preserve what was actually selected rather than assuming the visible quality label stayed fixed.
Separate user observation from instrumentation
A developer can use frame-related browser APIs to investigate composition timing, but those timestamps have specific meanings. A callback or request completion is not automatically the exact instant a viewer perceived the image on the physical panel. State the measurement method and its limit.
For a support record, the defined user action, endpoint, error messages and whether repeated starts behaved similarly can be enough to narrow the issue. Do not invent millisecond precision from a casual glance. The purpose is an accountable comparison that distinguishes a quickly displayed page asset from media that actually reached playback.
Questions about this guide
Is a poster the first decoded video frame?
No. It is a separate displayed asset.
Does one fast repeat establish universal startup speed?
No. Cache, connection and environment state affect the comparison.