Definition
Interaction to next paint (INP) measures the delay between a visitor interacting with a page and the browser painting the visual response, reported for a whole visit rather than for a single click. It covers taps, clicks and key presses, and it takes the slowest meaningful interaction of the visit as the score, which makes it a fair account of whether a page felt responsive throughout. Google places the good threshold at 200 milliseconds or less at the seventy-fifth percentile.
One slow filter tap, broken into its parts
- Input delay
- 120 milliseconds waiting for the main thread to become free, because an analytics script was still running
- Processing time
- 210 milliseconds of filter logic recalculating the whole product grid in one go
- Presentation delay
- 150 milliseconds for the browser to lay out and paint the updated grid
- Total for this interaction
- 480 milliseconds, which becomes the page's score if it is the slowest of the visit
- Good threshold
- 200 milliseconds or less
- Where the time went
- Two thirds of it in script the store controls, and none of it in the network
- After the work was split
- The grid updates in smaller pieces and the response is drawn before the recalculation has finished
Illustrative timings in the shape a browser profile reports them. The value of the breakdown is that each of the three parts has a different fix, and only a measurement says which one is costing the time.
Why it matters
INP matters to a store because it measures the moments a shopper is actively trying to do something: opening a size selector, applying a filter, adding to cart, opening the cart drawer. Those are the interactions closest to a purchase, and a page that takes half a second to acknowledge a tap invites a second tap, a duplicate action, or an exit. It replaced First Input Delay in Google's Core Web Vitals in March 2024, and the change matters for ecommerce because the old metric only looked at the first interaction, which on most stores was a harmless one. INP looks at the interactions people make while shopping.
Where it goes wrong
- Blaming the network, when the score is dominated by script execution and layout work happening on the visitor's own device
- Testing on a fast laptop, where main thread work that takes half a second on a mid-range phone completes quickly enough to hide the problem entirely
- Measuring the home page only, when the slow interactions live on category filters, variant pickers and the cart drawer
- Adding a loading spinner and calling it fixed, since the metric measures the paint that answers the interaction and a spinner is not the answer the shopper asked for
- Letting third-party scripts run in long uninterrupted blocks, which occupies the main thread at exactly the moment a shopper is tapping something
Questions about INP
What replaced First Input Delay and why?
Interaction to next paint replaced First Input Delay as a Core Web Vital in March 2024. The older metric measured only the delay before the browser started handling the first interaction, which flattered pages that responded quickly once and then became sluggish. INP looks across the visit and includes the work of producing the visual response, which is much closer to what a shopper experiences while browsing a catalogue.
Which ecommerce interactions most often cause a poor INP?
Category filters that rebuild a whole grid, variant selectors that re-render a product page on every change, cart drawers that fetch and lay out contents synchronously, and search suggestions that run on every keystroke. They share a pattern: a lot of work performed in one block in response to a single tap, on a device with far less processing headroom than the machine it was built on.
How is INP measured for a real store?
From field data collected on actual visits, which is what Search Console and the Core Web Vitals reporting tools show, aggregated at the seventy-fifth percentile and split by device. For diagnosis, a browser profile on a throttled mobile setting reproduces the slow interaction and shows where the milliseconds go. Use the field data to decide whether there is a problem and the profile to find its cause.