Field Data Versus Lab Data
The difference between what real users on real devices experience and what a synthetic run measures, and why the second is systematically optimistic.
Laboratory measurement runs a page on chosen hardware over a chosen connection at a chosen moment. It is reproducible, it attributes regressions to specific changes, and it can run before release — all valuable. It also measures one configuration, usually the one closest to the developer's.
Field data is what actual users experienced: their devices, their networks, their browser versions, their extensions, their cache state, their geography. It is the only measurement that answers whether the site is fast for the people using it.
The gap between them is consistently large and consistently in one direction. Distributions in the field have long tails — the 75th percentile of real users is frequently several times the laboratory number — because real traffic includes older devices, congested networks and cold caches, none of which a synthetic run reproduces by default.
The correct use is both, for different purposes: field data to know whether there is a problem and for whom, laboratory data to diagnose and to prevent regression. Reporting only laboratory numbers produces a dashboard that is green while customers on mid-range phones are waiting six seconds.