INP Replaces FID in March: The Interactivity Metric Gets Honest

The year closes with a scoreboard change that has a date attached, which is our favorite kind. Next March, the 12th, per the announced schedule, Interaction to Next Paint replaces First Input Delay as the responsiveness metric among the Core Web Vitals. The transition was announced in 2022, formalized this May, and has been surfacing in the tooling all year.

The honest translation, before the March coverage blurs it: FID graded on a curve. It measured only the input delay, the wait before the browser could begin processing your tap. A page could acknowledge the click instantly, then grind through half a second of JavaScript before anything visibly happened, and pass. Nearly everyone passed FID. The field-data pass rates ran embarrassingly high, while users demonstrably still waited on sluggish pages.

INP measures the full transaction. From input to the next paint, the visual response the user actually experiences. Across all interactions in the session rather than just the first. With the worst cases weighted: the reported value is effectively the session’s near-worst interaction. And with a 200-millisecond good threshold that reflects what responsiveness actually feels like. The consequence, visible already in the field data to anyone who looks: many currently green sites are about to be yellow. The quarter of runway between now and the March date is exactly the width our enforcement category exists to use.

Where the new failures will concentrate

The reconnaissance findings, from running INP audits across our book this fall. The failure geography is consistent, and it is not FID’s. Long JavaScript tasks after the click, above everything. The handler that does too much at once. The filter that re-renders the entire product grid. The menu that recalculates layout. The form that validates the world on every keystroke. The third-party tags competing for the main thread at interaction time, now metered per interaction. The analytics and personalization work riding input events.

Readers of the cost-of-JavaScript series will recognize every item, because INP is that series arriving at its enforcement metric. The main-thread bill we have itemized since 2017, finally presented per interaction, worst cases first.

And hydration weight in framework-heavy builds, the second concentration. The meta-framework tier’s known tax: the server-rendered page that looks ready while the framework re-attaches its event handlers, leaving early taps slow or dead. It lands directly on INP, exactly where our meta-framework coverage priced the risk in 2021. The sites that took that advice seriously (the budgets in CI, the hydration strategies, the islands patterns) will grade measurably better than the ones that took the defaults and drifted.

The fixes, correspondingly. Break up long tasks: yield to the main thread. The scheduling APIs are mature. Audit input handlers: what actually runs between tap and paint, profiled, not assumed. Defer the non-essential: the analytics call does not need to precede the visual response. And interrogate hydration: what re-attaches when, and whether the framework’s newer partial-hydration options apply. All standard practice in the tooling. None of it exotic. All of it exactly the work the old metric never made anyone do.

The pre-deadline audit, available now

• Read the INP data that already exists: CrUX has reported INP for over a year, and Search Console’s Core Web Vitals report already surfaces it. The pre-deadline audit is available to anyone who looks. Pull the field data by template cluster today, and know your March grade a quarter early.

• Prioritize by traffic-weighted worst templates, per the standing method. The interaction-heavy surfaces (search results, filtered listings, forms, mobile menus) are where INP failures cluster and where the user experience the metric approximates actually lives.

• Profile before optimizing: the long-task view in the dev tooling identifies the actual offender (first-party handler, third-party tag, hydration) in minutes. The fix differs entirely by which. The sites that skip the profiling will optimize the wrong tier.

• Annotate and re-baseline for March. The metric swap marked in analytics, the INP baselines recorded now, so the transition arrives as a new column in an existing dashboard, which is what the operations approach exists to make it.

The consolidation point, closing the program’s fourth year

The wider note this scoreboard change earns, as the vitals program closes its fourth year. Metrics will keep evolving toward user truth, and this transition is the pattern’s cleanest demonstration yet. FID was the honest best available at the program’s 2020 launch. Its generosity became legible as the field data piled up. The replacement measures closer to what users actually experience. With notice, tooling, and a date, per the program’s now-standard courtesy.

There will be successors to the successors. The measurement community’s work is public, and the practical advice is to read it annually rather than be surprised by it. The sites for whom each swap is routine are the ones we described in 2022, when monitoring converted from events to operations. Field monitoring running continuously. Budgets enforced in CI. Regression alerts staffed. A metric transition arrives as a column change rather than a crisis.

March will sort the programs from the projects, exactly as every enforcement date before it has sorted its cohort. The sorting test is the same one it has been since 2010: whether performance is something the organization does or something it did once. On schedule, as always. The announcement came in May, the data has been public for a year, and the date is March 12. This is the reminder that arrives while the runway still has slack. Use the quarter. The tap deserves its paint.

Closing out 2023

And the year-end paragraph, in the tradition, for the year the questions changed. 2023 opened with a boring Drupal upgrade and closed with copilots in every CMS. In between, search’s terrain was formally contested in February. The answer engine reached the results page in May. The crawler knocked in August. The update weather turned strange enough to demand candor in September. And the recurring topics (the accessibility bar, the vitals scoreboard, the platform anniversaries) kept their appointments regardless, which is precisely the point of keeping them.

The method held through all of it. Instruments first. Positions in writing. Calm as a competitive advantage. The owned spine under everything. And the method is what we carry into a year that will visibly demand it.

INP in March. The answer engine’s immigration status pending. The access-governance era opening. The copilots configured or not. The calendar is already interesting. See you in January, baselined.