Calculate two different rates
In a hypothetical request, a 1,000,000-byte body begins after a two-second wait and finishes one second later. The body-only interval suggests 8 Mb/s. Counting the full three-second request gives about 2.67 Mb/s for that object's completion interval. Both calculations use the same bytes and different denominators.
Neither is automatically the advertised broadband speed. The first excludes waiting; the second includes request-stage delays. Label them precisely so a developer does not compare a body-only rate in one run with a whole-request rate in another and claim an unexplained network change.
Keep the stages that the player needed
Browser resource timing can expose named stages such as request start and response start or end, subject to access rules and environment. Use the values according to their definitions. Name lookup, connection setup, server waiting and body receipt should not be invented from a single total.
For playback readiness, the resource's actual completion relative to media consumption matters. A segment that transfers rapidly after a long wait can still arrive too late for the available reserve. A high body-only rate therefore cannot alone explain away an observed interruption.
Report a comparable measurement
Retain the resource class, byte convention, timing boundaries, browser context and observation period. State whether the connection and cache were already warm. A cached response and a new remote fetch do not have equivalent setup costs, even if the inspector shows the same resource name.
If cross-origin timing restrictions hide details, report them as unavailable rather than filling in zero as a measured stage duration. Keep the actual user-visible start result beside instrumentation. The useful outcome is a transparent calculation that another person can reproduce, not a single speed number presented as proof of a healthy service or a precise attribution of every millisecond.
Questions about this guide
Can body transfer be quick after a long wait?
Yes. The stages can have very different durations.
Should hidden timing fields be treated as measured zeros?
No. Preserve the measurement limitation.