Custom Software · Performance · Production · Observability

Three thresholds, one verdict: what a fast first byte doesn't say about INP

Published · Updated

A server-rendered platform can hit a first byte under 200 ms and still fail Core Web Vitals. Passing requires all three metrics at the 75th percentile, and rendering choice only speaks to some of them.

On the Abbys Consult platform, time to first byte came in at under 200 ms. That number is easy to quote and easy to misread. It describes how quickly a response starts arriving, not whether the page passes the measurement that tools actually report. The verdict on page experience is assembled from three metrics at one percentile, and a fast first byte is an input to part of it rather than a substitute for the whole.

The verdict is three numbers at one percentile

Web Vitals sets explicit targets: LCP should occur within 2.5 seconds of when the page first starts loading, and pages should have an INP of 200 milliseconds or less. The recommended measurement point is the 75th percentile of page loads, segmented across mobile and desktop devices. Tools assessing compliance should consider a page passing if it meets the recommended targets at the 75th percentile for all three Core Web Vitals metrics.

Two things follow from that wording. First, the unit of judgement is a percentile of real page loads, not a single run on one machine. Second, the metrics are not scored as an average you can offset. Meeting one target generously does not buy slack on another. All three have to clear their threshold at the same percentile, on each device class you serve.

What server rendering is documented to buy

TTFB measures the time between starting navigation to a page and when the first byte of a response begins to arrive. Good values are 0.8 seconds or less, and poor values are greater than 1.8 seconds. It is explicitly not a Core Web Vitals metric, so it is not absolutely necessary to meet the good threshold, provided that doing so does not impede the ability to score well on the metrics that matter.

The documented trade is specific: a server-rendered site that does not require as much client-side work could have a higher TTFB but better FCP and LCP values than an entirely client-rendered experience. That is a statement about the load path. A first byte under 200 ms sits comfortably inside the good band and supports the LCP side of the verdict. Nothing in that trade is stated to move INP.

INP needs its own budget

This is where a fast platform quietly fails. INP has its own target of 200 milliseconds or less, measured at the same 75th percentile, and because passing requires all three metrics to clear their thresholds, an INP failure sinks the verdict however good the first byte was. Treat interaction responsiveness as a separate budget with a separate owner, not as something the rendering decision already handled.

Budget it the way the threshold is written. Collect INP from real page loads rather than a single lab run, keep the mobile and desktop segments apart, and look at the 75th percentile rather than the median. A developer machine on a wired connection is not the percentile that gets reported, and a median that looks fine can hide a quarter of loads sitting above the target.

Percentiles multiply with deploy targets

The Tatano Energy platform runs one codebase across four country domains indexed separately, serving seven languages. Each of those is a place where real page loads happen, and the measurement guidance is already per device class. An aggregate across all of them can clear every threshold while one market sits above the INP or LCP target. Decide up front which segments you report, then report them individually.

That also changes what to ask for when buying rather than building. A single composite score tells you almost nothing about this verdict. Ask for LCP and INP at the 75th percentile, split by mobile and desktop, for each domain in production, plus TTFB as a diagnostic alongside them rather than in place of them. If a supplier can only show one number, they are not measuring what the tools report.

The pragmatic version: a first byte under 200 ms is a real result and worth having, but it is evidence about one part of the load path. Keep TTFB as a diagnostic, keep LCP and INP as the thresholds you have to clear, measure both at the 75th percentile per device class and per domain, and give interaction responsiveness its own budget before the rendering choice gets credit for it.

Sources

web.dev (Google Chrome) — Web Vitals — https://web.dev/articles/vitals

web.dev (Google Chrome) — Time to First Byte (TTFB) — https://web.dev/articles/ttfb

Neurolinks case study — A consulting firm, dressed for trust — https://neurolinks.be/work/abbys-consult

Neurolinks case study — Four markets, one codebase — https://neurolinks.be/work/tatano-energy

Working on a project where these methods apply?