Interaction to Next Paint became the metric that speaks in revenue. The LCP asked how fast the page appears, and INP asks how fast the page answers, and the enterprise page that hesitates leaks the interaction.
The field data shows the pattern. The pages with the sub-200 ms INP keep the user engaged, and the pages that cross the 300 ms line see the abandonment and the conversion drift the business feels.
This guide covers the INP mechanics, the fixable causes and the budget that ships without a full rebuild.
Why INP Is the Revenue Metric
INP measures the response to the interaction. The click, the tap and the keypress race the render, and the delay that follows the user's intent is the delay they feel.
The long tasks are the culprit. The main thread that parses, executes and renders in one 250 ms block makes every interaction wait behind it.
The field correlation is consistent. The slow-INP pages show the higher bounce on the interactive pages, and the checkout and the filter-heavy flows suffer first.
| INP | Field Quality | Revenue Pattern |
|---|---|---|
| Under 200 ms | Good | Interaction held |
| 200-400 ms | Needs improvement | Visible drift |
| Over 400 ms | Poor | Abandonment climbs |
The Fixable Causes
The render-blocking script is first. The bundle that parses on the critical path delays the first interaction-ready state, and the deferral frees the thread.
The oversized event is second. The handler that re-renders the whole screen, hits every listener and blocks for layout decays the responsiveness the user expects.
The third-party work is third. The widget that initializes on load, fights the main thread and ships its own long task pushes the poor field data the lab never measures.
The Fix List That Ships
The 100 ms budget starts in the lab. The throttled trace identifies the long tasks, and the attribution names the script, the handler or the widget responsible.
The script strategy buys the thread. The defer, the code splitting and the critically rendered path reduce the parse work, and the interaction moves to the front of the queue.
The passive handlers and the reduced listeners trim the event cost. The touch-action, the requestAnimationFrame batching and the layout Only writes keep each handler cheap.
Vitals Failures
The lab-only score is first. A page measured on the empty desktop profile hides the production scripts and the real field data, and the release ships the regression.
The wallpaper tweak is second. A font swap that brightens the LCP while the thread stays blocked moves the score without moving the feel, and the user keeps waiting.
The one-shot audit is third. A fix applied once while the third-party widget updates quarterly undoes the thread, and the vitals decay without the monitoring.
- Scores measured on lab-only profiles
- Fixes that polish the number, not the feel
- One-shot audits without the field watching
- Third-party updates silently reshipping the work
The Budget That Holds
Watch the field percentiles. The site's real INP at the 75th percentile is the number the revenue feels, and the lab is the debugger, not the report.
Package the fix list to ship. The deferred scripts, the slimmed handlers and the held third-party long task go out in the release, and the field data follows the release.
Keep the budget owned. The monthly trace, the release-time comparison and the third-party review keep the thread free, and the responsiveness stays a standing discipline.
Implementation Checklist
- Track the field INP at the 75th percentile
- Attribute the long tasks in a throttled trace
- Defer render-blocking scripts
- Trim the event listeners and the repaints
- Shard or defer the third-party long tasks
- Compare the field data after every release
Final Thoughts
INP is the Core Web Vital that speaks in revenue. The thread that responds to the interaction keeps the flow alive, and the deferred work that frees it ships with the field proof.
Measure the field, fix the thread and keep the budget owned. The responsiveness holds because the squad watches the percentile, not the wallpaper.
Frequently Asked Questions
LCP asks how fast the page appears, and INP asks how fast it answers. The interactive flows, the checkout and the filters hinge on the response, and the revenue drift follows the delay.
The long tasks on the main thread. Render-blocking scripts, heavy event handlers and third-party widgets that initialize on load all queue the interaction behind them.
Defer the scripts, split the long tasks, slim the handlers and shard the third-party work. Each is a targeted release, and the field percentiles confirm the gain.
