暫定日本語訳：First Input Delay (FID)

暫定日本語案: TL;DR — First Input Delay (FID) is an old metric. It measured how long your 暫定日本語案: ページ made someone wait before it could even start reacting to their first click 暫定日本語案: or tap. Google retired it in March 2024 and replaced it with INP, then removed it 暫定日本語案: from its tools entirely in September 2024. So there’s nothing to fix here anymore — 暫定日本語案: but it’s worth knowing what it was if you run into “FID” in old reports.

暫定日本語案: What First Input Delay was

暫定日本語案: When you tap a button and nothing happens for a beat, the ページ feels broken — even 暫定日本語案: if it looks loaded. First Input Delay (FID) was Google’s way of putting a 暫定日本語案: number on that specific frustration.

暫定日本語案: FID measured one narrow thing: the gap between your first interaction with a 暫定日本語案: ページ (a click, a tap, a key press) and the moment the browser was actually free 暫定日本語案: to start responding to it. If the browser was busy running JavaScript when you 暫定日本語案: tapped, your tap had to wait in line. That wait was the “delay.”

暫定日本語案: Two things it did not measure:

• 暫定日本語案: How long the button’s code took to run once it started.
• 暫定日本語案: How long the ページ took to visually update afterward.

暫定日本語案: Just the wait before anything could begin. That narrowness is a big reason it 暫定日本語案: eventually got replaced.

暫定日本語案: What counted as a good score

暫定日本語案: FID was scored in milliseconds:

• 暫定日本語案: Good: 100 ms or less
• 暫定日本語案: Needs improvement: 100–300 ms
• 暫定日本語案: Poor: over 300 ms

暫定日本語案: Why you don’t need to worry about it anymore

暫定日本語案: Here’s the important part for anyone reading this in 2026: FID is retired. 暫定日本語案: Google replaced it as a Core Web Vital with INP (Interaction to Next Paint) on 暫定日本語案: March 12, 2024 — 検索 Console stopped showing FID that same day. PageSpeed 暫定日本語案: Insights and the CrUX API kept reporting it a little longer, then dropped it on 暫定日本語案: September 9, 2024. If a tutorial or an old dashboard still lists FID as a current 暫定日本語案: Core Web Vital, that コンテンツ is out of date.

暫定日本語案: INP does the same job better: instead of only timing the first interaction’s 暫定日本語案: wait, it measures the full responsiveness of every interaction throughout a 暫定日本語案: visit. If you’re trying to make your サイト feel snappy today, INP is the metric to 暫定日本語案: watch, not FID.

暫定日本語案: Want the full history — the exact thresholds, why FID was designed so narrowly, 暫定日本語案: where old FID データ still lives, and how it maps onto INP — switch to the 暫定日本語案: Advanced tab.

暫定日本語案: TL;DR — FID was the Core Web Vital for responsiveness until INP replaced it on 暫定日本語案: March 12, 2024 — the date 検索 Console also dropped it from its report. Chrome 暫定日本語案: tools, PageSpeed Insights, and the live CrUX API kept it a little longer and cut it 暫定日本語案: on September 9, 2024, each on its own schedule. It measured only the input 暫定日本語案: delay of the first interaction — not handler runtime, not repaint — deliberately, 暫定日本語案: to avoid perverse incentives. Thresholds: good ≤100 ms, poor >300 ms at p75, field 暫定日本語案: only (Total Blocking Time was the lab proxy — a correlated diagnostic, not a 暫定日本語案: conversion formula). Don’t confuse those thresholds with INP’s 200/500 ms, and 暫定日本語案: don’t try to convert one metric’s number into the other. Poor FID came from 暫定日本語案: main-thread contention — long JavaScript tasks — which is exactly what causes poor 暫定日本語案: INP and TBT, so the legacy fixes still pay off. Historical FID データ survives in 暫定日本語案: the CrUX BigQuery dataset (through the 202409 release); it’s gone everywhere 暫定日本語案: live.

暫定日本語案: What FID actually measured

暫定日本語案: Google’s definition was precise. Per web.dev: 暫定日本語案: “FID measures the time from when a user first interacts with a ページ (that is, when 暫定日本語案: they click a リンク, tap on a button, or use a custom, JavaScript-powered control) to 暫定日本語案: the time when the browser is actually able to begin processing event handlers in 暫定日本語案: response to that interaction.”

暫定日本語案: Read that carefully, because the scope is the whole story. FID captured the 暫定日本語案: delay before processing could begin — and nothing after. Not how long the event 暫定日本語案: handler ran. Not how long the ページ took to paint the result. Just the wait.

暫定日本語案: Why was the browser ever “not able to begin”? web.dev is blunt about the cause: 暫定日本語案: “In general, input delay (a.k.a. input latency) happens because the browser’s main 暫定日本語案: thread is busy doing something else, so it can’t (yet) respond to the user.” There 暫定日本語案: is one main thread, and if it’s mid-task parsing or executing JavaScript when the 暫定日本語案: user acts, the interaction sits in the queue until that task finishes. I make the 暫定日本語案: same point in my Ahrefs FID guide: 暫定日本語案: there’s just one main thread, JavaScript competes to run tasks on it, and while a 暫定日本語案: task is running the ページ can’t respond to input — that stall is the delay a user 暫定日本語案: actually feels.

暫定日本語案: Why FID only measured the delay (not the whole interaction)

暫定日本語案: This looks like a design flaw until you understand the reasoning. Google measured 暫定日本語案: only input delay on purpose. Folding the handler’s execution time and the repaint 暫定日本語案: into the metric could, as web.dev explains, incentivize developers to game it — 暫定日本語案: they could wrap their event-handler logic in an asynchronous callback to split it 暫定日本語案: off from the interaction’s task and make the number look better while the actual 暫定日本語案: experience got worse. So FID stayed narrow.

暫定日本語案: That narrowness is also FID’s fatal limitation. A ページ could post a great FID and 暫定日本語案: still feel sluggish, because every interaction after the first went unmeasured, 暫定日本語案: and the slow part of an interaction is often the processing and repaint FID ignored. 暫定日本語案: That gap is precisely what INP was built to close.

暫定日本語案: The thresholds — and the one you’ll see people get wrong

暫定日本語案: Measured at the 75th percentile of ページ loads, segmented across mobile and 暫定日本語案: desktop. web.dev’s guidance was simply that sites should strive for a First Input 暫定日本語案: Delay of 100 milliseconds or less. My own 暫定日本語案: FID article uses the same 暫定日本語案: figures — good ≤100 ms, needs improvement >100 ms and ≤300 ms, poor >300 ms.

暫定日本語案: The common mistake: confusing FID’s thresholds with INP’s. They are different 暫定日本語案: numbers for different metrics. FID = 100 ms good / 300 ms poor. INP = 200 ms good / 暫定日本語案: 500 ms poor. Several third-party summaries — and even automated コンテンツ passes — 暫定日本語案: conflate the two, so if you see “200 ms” cited as FID’s good threshold, it’s wrong.

暫定日本語案: FID was a field-only metric

暫定日本語案: You could never get FID from Lighthouse or any lab tool, because it required a real 暫定日本語案: user’s real first interaction — web.dev states plainly that FID is a metric that can 暫定日本語案: only be measured in the field, as it requires a real user to interact with your ページ. 暫定日本語案: Lab tools don’t click, so there was nothing for FID to time.

暫定日本語案: The lab stand-in was always Total Blocking Time (TBT). As I put it in 暫定日本語案: my PageSpeed Insights guide, you 暫定日本語案: won’t find FID or INP in lab データ — those require clicks on the ページ that lab 暫定日本語案: testing doesn’t reproduce — so you use Total Blocking Time as a proxy metric to work 暫定日本語案: on improving instead. That relationship outlived FID: TBT is now the lab proxy for 暫定日本語案: INP.

暫定日本語案: One guard worth stating plainly: TBT is a correlated diagnostic, not a conversion 暫定日本語案: formula. There was never an equation that turned a TBT number into an exact FID 暫定日本語案: number, and there isn’t one for INP either — a bad TBT score tells you main-thread 暫定日本語案: work is a likely culprit, not what your field FID or INP would have been.

暫定日本語案: Why FID was retired: the INP transition

暫定日本語案: FID’s replacement was announced well in advance. Per web.dev, INP officially became 暫定日本語案: a Core Web Vital and replaced FID on March 12, 2024, at which point FID was 暫定日本語案: deprecated and removed from the program. Google’s stated reasoning: over time it 暫定日本語案: became clear a new metric was needed to capture aspects of interactivity that FID 暫定日本語案: did not.

暫定日本語案: The timeline had two distinct milestones — worth keeping straight, because it’s 暫定日本語案: easy (and common, even in automated コンテンツ) to collapse them into a single date:

1. 暫定日本語案: March 12, 2024 — retired as a Core Web Vital. INP took over; FID was no 暫定日本語案: longer part of the ranking-relevant Core set. 検索 Console removed FID from 暫定日本語案: its Core Web Vitals report that same day.
2. 暫定日本語案: September 9, 2024 — removed from the tools, on 商品-specific schedules. 暫定日本語案: Per web.dev, as of that date FID was no longer supported in Chrome tools. 暫定日本語案: PageSpeed Insights stopped displaying real-user FID データ and the CrUX API 暫定日本語案: discontinued serving the metric going forward; the CrUX BigQuery dataset stopped 暫定日本語案: adding new FID fields starting with the 202409 release, though earlier months 暫定日本語案: stayed queryable.

暫定日本語案: It’s not accurate to say every Google surface dropped FID on September 9 — 検索 暫定日本語案: Console’s cutoff was six months earlier, tied to the INP replacement, not the later 暫定日本語案: tools cleanup.

暫定日本語案: web.dev’s own FID article now leads with the retirement notice: First Input Delay is 暫定日本語案: no longer a Core Web Vital, and has been replaced by the Interaction to Next Paint 暫定日本語案: (INP) metric. And Google 検索 Central’s current Core Web Vitals documentation 暫定日本語案: doesn’t mention FID at all — it covers only LCP, INP, and CLS. When the official 暫定日本語案: ranking doc stops naming a metric, that’s about as retired as it gets.

暫定日本語案: FID vs INP: what changed

暫定日本語案: The two metrics measure genuinely different things, which is why you can’t just map 暫定日本語案: one onto the other:

暫定日本語案: The through-line: FID timed the front door of one interaction; INP times the 暫定日本語案: whole journey of every interaction. There’s no formula that converts an old FID 暫定日本語案: number into an equivalent INP number, and a ページ’s FID relative to other ページ 暫定日本語案: doesn’t predict its INP ranking relative to those same ページ — they’re measuring 暫定日本語案: different interaction sets against different endpoints, so any resemblance between 暫定日本語案: the two numbers on a given ページ is coincidental, not a rule. See 暫定日本語案: Interaction to Next Paint 暫定日本語案: for the full treatment of the metric that replaced it.

暫定日本語案: Where old FID データ still lives

暫定日本語案: Retirement didn’t vaporize the historical record. What’s gone versus what remains:

• 暫定日本語案: Gone (live/current-facing): 検索 Console’s Core Web Vitals report dropped 暫定日本語案: FID on March 12, 2024, the day INP took over. The PageSpeed Insights UI and the 暫定日本語案: live CrUX API kept reporting it a little longer and stopped on September 9, 2024.
• 暫定日本語案: Still there (historical): FID データ from before the cutoff remains queryable 暫定日本語案: in the public CrUX BigQuery dataset — but only through the 202409 dataset; 暫定日本語案: BigQuery stopped adding new FID fields starting with that release, though earlier 暫定日本語案: months stayed in place. If you need to reconstruct a サイト’s old responsiveness 暫定日本語案: history, that’s where to look — not the live tools. Pin the dataset month when you 暫定日本語案: cite a number, label it historical, and don’t treat a legacy FID figure as 暫定日本語案: numerically comparable to a current INP figure — there’s no conversion between 暫定日本語案: them (see the comparison table above).

暫定日本語案: Does FID still matter today?

暫定日本語案: Directly, no — there’s nothing left to measure or report, so there’s nothing to 暫定日本語案: “fix.” But the causes of poor FID and the causes of poor INP are nearly 暫定日本語案: identical: long JavaScript tasks hogging the main thread. So any work you already 暫定日本語案: did to improve FID wasn’t wasted. As I note in 暫定日本語案: my FID guide, even though FID was 暫定日本語案: replaced by INP in March 2024, it’s still worth working on the same underlying 暫定日本語案: issues — many of the things you do to improve TBT and FID also improve INP.

暫定日本語案: The fixes that reduced FID are the same ones that help INP and TBT now:

• 暫定日本語案: Reduce the amount of JavaScript you ship.
• 暫定日本語案: Load JavaScript later where you can (async/defer).
• 暫定日本語案: Break up long tasks with code splitting so no single task monopolizes the 暫定日本語案: main thread.
• 暫定日本語案: Move work off the main thread with web workers.
• 暫定日本語案: Use server-side rendering or prerendering to cut client-side work.

暫定日本語案: Was FID ever a big ranking factor?

暫定日本語案: Even while active, FID — as part of Core Web Vitals — was never a heavy ranking 暫定日本語案: signal. Google representatives have repeatedly characterized Core Web Vitals as 暫定日本語案: closer to a tiebreaker than a primary signal, applied only when other things are 暫定日本語案: roughly equal. My own read, from 暫定日本語案: my Core Web Vitals guide, is the same: 暫定日本語案: “I don’t think Core Web Vitals have much impact on SEO and, unless you are 暫定日本語案: extremely slow, I generally won’t prioritize fixing them.” FID rarely got singled 暫定日本語案: out on its own in rep commentary — it was almost always discussed as part of the 暫定日本語案: Core Web Vitals bundle, not as a standalone ranking lever.

暫定日本語案: Bing and FID

暫定日本語案: There’s effectively no Bing-specific angle here. Bing never adopted Core Web Vitals 暫定日本語案: as a named ranking signal the way Google did, and it never published FID (or INP) 暫定日本語案: thresholds of its own. Bing cares about fast, responsive ページ in general terms, but 暫定日本語案: FID was a Google-ecosystem metric from start to finish.

暫定日本語案: Where to go next

暫定日本語案: FID sits under the Web Vitals 暫定日本語案: initiative, in the Web パフォーマンス cluster. The 暫定日本語案: metrics most relevant to FID:

• 暫定日本語案: Interaction to Next Paint — the Core Web Vital that replaced it, and the one 暫定日本語案: to actually optimize for now.
• 暫定日本語案: Total Blocking Time — the lab proxy that stood in for FID (and now stands in 暫定日本語案: for INP) whenever you couldn’t measure real interactions.
• 暫定日本語案: Core Web Vitals — the ranking-relevant trio (LCP, INP, CLS) FID used to belong 暫定日本語案: to.

暫定日本語案: AI summary

暫定日本語案: A condensed take on the Advanced version:

• 暫定日本語案: FID = a retired Core Web Vital for responsiveness. It measured only the 暫定日本語案: input delay of the first interaction — the wait before the browser could 暫定日本語案: begin processing the event handler — not the handler’s runtime or the repaint.
• 暫定日本語案: Thresholds: good ≤ 100 ms, needs improvement 100–300 ms, poor > 300 ms, at the 暫定日本語案: 75th percentile, mobile/desktop split. Field only — never measurable in 暫定日本語案: Lighthouse; Total Blocking Time was the lab proxy.
• 暫定日本語案: Don’t confuse thresholds: FID = 100/300 ms; INP = 200/500 ms. Different 暫定日本語案: metrics, different numbers.
• 暫定日本語案: Cause of poor FID: main-thread contention — long JavaScript tasks. Same root 暫定日本語案: cause as poor INP and TBT.
• 暫定日本語案: Retirement timeline: replaced by INP on March 12, 2024 — the same day 暫定日本語案: 検索 Console dropped FID from its report. Chrome tools, PageSpeed Insights, 暫定日本語案: and the live CrUX API stopped supporting it on September 9, 2024, each on its 暫定日本語案: own schedule — not a single universal cutoff. Google 検索 Central’s current CWV 暫定日本語案: doc no longer mentions it.
• 暫定日本語案: FID vs INP: FID timed the first interaction’s delay; INP measures all 暫定日本語案: interactions’ full latency (delay + processing + presentation). No formula 暫定日本語案: converts one metric’s number into the other.
• 暫定日本語案: Old データ: gone from the live tools; pre-cutoff FID データ still lives in the 暫定日本語案: CrUX BigQuery dataset through the 202409 release — pin the dataset month and 暫定日本語案: don’t compare it numerically to current INP.
• 暫定日本語案: No conversion, ever: TBT is a correlated lab proxy for FID and INP, not a 暫定日本語案: formula that translates one into the other.
• 暫定日本語案: Still matter? Nothing to fix directly, but the JS fixes (reduce/defer JS, 暫定日本語案: break up long tasks, web workers, SSR) carry straight over to INP and TBT.
• 暫定日本語案: Ranking weight: minor even when active — CWV framed as a tiebreaker; Patrick: 暫定日本語案: don’t prioritize unless you’re extremely slow. Bing has no FID equivalent.

暫定日本語案: Official documentation

暫定日本語案: Primary-source documentation on FID and its retirement.

暫定日本語案: Google / web.dev

• 暫定日本語案: First Input Delay (FID) — the metric’s definition, thresholds, why only input delay was measured, and the retirement notice (Philip Walton; updated 2024-10-06).
• 暫定日本語案: Chrome ends support for First Input Delay — the September 2024 removal from Chrome tools, PSI, and the CrUX API (Rick Viscomi).
• 暫定日本語案: Interaction to Next Paint becomes a Core Web Vital on March 12 — the announcement that INP would replace FID (Jeremy Wagner, Rick Viscomi).
• 暫定日本語案: Understanding Core Web Vitals and Google 検索 results — the current ranking doc, which now lists only LCP, INP, and CLS (no FID).
• 暫定日本語案: Interaction to Next Paint (INP) — the metric that replaced FID.

暫定日本語案: MDN

• 暫定日本語案: First Input Delay (FID) glossary — a short reference definition.

暫定日本語案: Bing / Microsoft

• 暫定日本語案: None specific to FID. Bing never published FID thresholds or named Core Web Vitals as a ranking signal of its own.

暫定日本語案: Quotes from the source

暫定日本語案: On-the-record statements from Google’s official web.dev documentation. Each リンク is 暫定日本語案: a deep リンク that jumps to the quoted passage on the source ページ.

暫定日本語案: Google / web.dev — what FID measured

• 暫定日本語案: “FID measures the time from when a user first interacts with a ページ (that is, when they click a リンク, tap on a button, or use a custom, JavaScript-powered control) to the time when the browser is actually able to begin processing event handlers in response to that interaction.” 暫定日本語案: Jump to quote
• 暫定日本語案: “In general, input delay (a.k.a. input latency) happens because the browser’s main thread is busy doing something else, so it can’t (yet) respond to the user.” 暫定日本語案: Jump to quote

暫定日本語案: Google / web.dev — the “good” threshold

• 暫定日本語案: “To provide a good user experience, sites should strive to have a First Input Delay of 100 milliseconds or less.” 暫定日本語案: Jump to quote

暫定日本語案: Note: web.dev renders some コンテンツ via JavaScript, so the #:~:text= deep 暫定日本語案: リンク may need confirmation against the live ページ. The retirement details (March 12, 暫定日本語案: 2024 replacement; September 9, 2024 tool removal), the field-only nature of FID, the 暫定日本語案: design rationale for measuring only input delay, and Google reps’ “tiebreaker” 暫定日本語案: characterization of Core Web Vitals are stated in Google’s docs and blog posts but 暫定日本語案: are paraphrased here rather than quoted verbatim, because the source ページ were not 暫定日本語案: independently re-fetched for exact wording. Patrick’s lines from his Ahrefs guides 暫定日本語案: are reproduced as his own words with the exception of any threshold figure flagged as 暫定日本語案: possibly conflated at research time — the authoritative Patrick-sourced numbers are 暫定日本語案: the 100/300 ms figures from his FID article.

暫定日本語案: First Input Delay cheat sheet

暫定日本語案: Status: retired. Keep this for reading legacy データ and old reports — not as an 暫定日本語案: optimization target.

暫定日本語案: FID at a glance

暫定日本語案: FID vs INP — don’t mix them up

暫定日本語案: Key dates

• 暫定日本語案: March 12, 2024 — INP replaces FID as a Core Web Vital; 検索 Console drops 暫定日本語案: FID from its report the same day.
• 暫定日本語案: September 9, 2024 — Chrome tools, PageSpeed Insights, and the live CrUX API 暫定日本語案: stop supporting FID, on their own schedule (not the same date as GSC).

暫定日本語案: Where FID データ is now

• 暫定日本語案: Live tools (PSI UI, GSC, CrUX API): gone (GSC since March 12, 2024; PSI/CrUX 暫定日本語案: since September 9, 2024).
• 暫定日本語案: Pre-cutoff history: CrUX BigQuery dataset, through the 202409 release. Don’t 暫定日本語案: compare its numbers directly to current INP — there’s no conversion.

暫定日本語案: Fixing the underlying issue (helps INP/TBT now)

• 暫定日本語案: Reduce JavaScript · defer/async load · break up long tasks (code-split) · web 暫定日本語案: workers · SSR/prerendering.

暫定日本語案: Mistakes to avoid with legacy FID データ

暫定日本語案: Treating FID as a current Core Web Vital

暫定日本語案: FID was replaced by INP in March 2024 and removed from current Chrome reporting 暫定日本語案: surfaces later that year. Use INP for current responsiveness work; keep FID only when 暫定日本語案: interpreting historical datasets.

暫定日本語案: Comparing FID and INP against the same thresholds

暫定日本語案: FID’s historical Good/Poor thresholds were 100/300 ms; INP’s are 200/500 ms. The 暫定日本語案: numbers are not interchangeable because FID measured only pre-handler delay, while 暫定日本語案: INP covers the interaction through the next paint.

暫定日本語案: Looking for FID in a lab test

暫定日本語案: FID required a real first user input and was field-only. Total Blocking Time was the 暫定日本語案: lab proxy; a Lighthouse result was never a direct FID observation.

暫定日本語案: Optimizing a retired score instead of the user problem

暫定日本語案: Do not preserve a FID dashboard as the goal. Break up long main-thread tasks and 暫定日本語案: reduce blocking JavaScript, then measure current responsiveness with INP in the field.

暫定日本語案: Test yourself: First Input Delay (FID)

暫定日本語案: Five quick questions on what FID measured and why it’s retired. Pick an answer for 暫定日本語案: each, then check.

変更履歴