Get NitroPack with up to 37% OFF
Interaction to Next Paint (INP) is now an official Core Web Vital, replacing First Input Delay (FID) as the new responsiveness metric.
All FID improvements you've previously made will serve as a great foundation. However, INP introduces new performance challenges, and in the following lines, you will learn exactly how to deal with them in order to provide your visitors with an exceptional user experience.
Read on.
INP assesses a page's overall responsiveness to user interactions by observing the latency of all qualifying interactions during a user's visit to a page. The final INP value is the longest interaction observed.
The interactions that play a part in INP’s calculations are:
Similar to the other Core Web Vitals, your INP score can be in one of the three thresholds:
To guarantee that you achieve this objective for the majority of your users, it is recommended to assess the 75th percentile of page loads, segmented across mobile and desktop devices.
If you want to learn more or brush up your knowledge about INP, read our article on the new responsiveness metric.
If you want your INP score to go from poor to good, you need to understand interaction latency.
So what exactly is interaction latency?
Interaction latency refers to the delay or lag experienced between a user's input or action and the resulting response or output on the screen. It is a crucial factor in determining your site's responsiveness and perceived performance.
Three primary components contribute to interaction latency:
Input delay refers to the time between when a user starts interacting with the page and when the associated actions or event callbacks begin to execute. It includes the physical or technical delays caused by the input device (e.g., keyboard, mouse, touchscreen) and the system's input processing mechanisms.
Once the user input is received, the system must process it to determine the appropriate response or action. Processing time refers to the duration required for the system to analyze and interpret the input data, perform any necessary calculations or operations, and generate the output or response.
After the system has generated the output or response, there is typically a delay before it is presented to the user. Presentation delay encompasses the time it takes for the system to update the display, render graphics or user interfaces, and deliver the output to the user interface or output device.
If you need more information, you can check Jeremy Wagner’s presentation at JSConf Korea 2022:
Understanding and optimizing the interaction latency can provide a seamless user experience and fix your INP scores.
But before that, let’s take a look at the main culprits for high interaction latency and bad INP scores…
Although INP is labeled as pending, this does not mean you should enter the optimization process without a strategy.
The first thing you need to do is learn what the main INP culprits are, so you can handle them effectively.
Here are the main reasons for the “INP issue: longer than 200ms” error message:
Everything that a browser does is considered a task. This includes rendering, parsing HTML, running JavaScript, and anything you may or may not have control over.
The main thread is where the browser does most of the work needed to display a page. And while there might be dozens of tasks that need to be executed, the main thread can only process one task at a time.
But that’s only half of the problem.
The other half is that if a task takes more than 50 milliseconds to be executed, it’s classified as a long task.
Considering that the main thread can handle one task at a time, the longer the task is, the longer the browser will be blocked processing it.
In other words, if the user is attempting to interact with the page while a long task runs, the browser will be delayed in fulfilling the request.
The result is - interaction latency and a lower INP score.
Another reason for failing INP is having a large DOM size.
The Document Object Model (DOM) is an inseparable part of every web page. The DOM is a representation of an HTML document structured as a tree. Each branch in the tree terminates at a node, and each node contains objects. Nodes can represent different parts of the document, such as elements, text strings, or comments.
The DOM itself isn’t a problem, but its size might be. A large DOM size impacts a browser's ability to render a page quickly and efficiently.
The larger the DOM, the more resource-intensive it is to initially display the page and make subsequent updates during the page's lifecycle.
Simply put:
If you want a page to respond quickly to user interactions, ensure your DOM includes only the necessary elements.
You might be wondering what “necessary” means. According to Lighthouse, a page's DOM size is excessive when it exceeds 1,400 nodes.
So make sure to stay within this limit. Otherwise, you might see the following error in your PageSpeed Insights report:
To understand why client-site rendering might cause poor INP scores, we need to explain how it differs from the server-side rendering of HTML.
The traditional page load involves the browser receiving HTML from the server on every navigation. What happens in the background when a person decides to load a page is:
The key here is “in chunks.”
When the browser receives the first chunk of HTML, it can start parsing it. But as we know, parsing HTML is a task the main thread handles.
However, after each chunk is processed, the browser takes a break from parsing and allows other tasks to be performed. This prevents long tasks that could slow down the browser.
Instead, it can start rendering the parts of the page that have already been parsed, so the user sees a partially loaded page sooner. It can also handle any user interactions that occur during the initial loading of the page.
In other words:
This approach translates to a better Interaction to Next Paint (INP) score for the page.
On the contrary, if your website uses the single page application (SPA) pattern, which dynamically creates large parts of the HTML/DOM on the client with JavaScript, you can expect negative effects on your INP score.
In client-side rendering, the server sends a small chunk of basic HTML to the client. Then the client takes care of filling in the main content of the page using data it fetches from the server.
All future navigations are handled by JavaScript, fetching new HTML from the server and dynamically updating the page without fully reloading it. Unfortunately, when it comes to JavaScript tasks on the client side, they are not automatically chunked up.
This can lead to long tasks that block the main thread, potentially affecting your page's Interaction to Next Paint score. Therefore, client-side rendering can hurt the loading and interactivity of your page.
If you need additional info on the pros and cons of server-side vs. client-side rendering, check out this video:
Now that you know some of the main culprits, let’s proceed with measuring your INP score and identifying slow interactions.
The next step in the INP optimization journey is to measure your site's performance and identify any slow interactions.
Similar to First Input Delay, INP is best measured in the field – examining how real users experience your website.
The optimal testing process would look like this:
We said optimal because, in some cases, your site might not have field data (also known as Real User Monitoring (RUM) data). However, this doesn’t mean you should give up optimizing your INP score. You need to take a different approach and leverage the available lab tools.
Let’s take a look at both scenarios and explain how to take most of your field and lab data.
Ideally, you'll want to have field data when you start improving your site’s responsiveness. Relying on RUM data saves you a lot of time figuring out which interactions need to be optimized.
Furthermore, lab-based tools can simulate certain interactions but cannot fully replicate real-world user experiences.
When gathering INP field data, you'll want to capture the following to give context to why interactions were slow:
If you’re asking yourself:
How am I supposed to capture all of these things?
Don’t worry. All of the data is exposed in the web-vitals JavaScript library. You can check Google’s step-by-step guide on how to leverage the web-vital library and even how to transmit INP data straight to your Google Analytics.
Also, even if you’re already collecting data with a third-party RUM provider, consider comparing it with Chrome UX Report (CrUX) data, as there are differences in the methodologies they use.
Field data is the most reliable source for measurement. However, as we said, it is not always available.
But there’s no need to worry because you can still measure and identify interactions to improve with the help of lab data.
You can use either Lighthouse or PageSpeed Insights to run some performance tests. The metric you should keep an eye out for is Total Blocking Time (TBT).
TBT is a metric that assesses page responsiveness during load and correlates very well with INP. A poor TBT score is a signal there are interactions that might be slow during page load.
Here’s how you can reproduce slow interaction with lab tools:
The Web Vitals Chrome Extension is one of the easiest ways to measure your site’s interaction latency. Here’s what you need to do to retrieve useful information:
Finally, open the Chrome DevTools console and begin testing. You'll receive helpful console logs giving you detailed diagnostic information for the interaction.
To get even more information about why interaction is slow, you can use the performance profiler in Chrome DevTools. Just do the following:
To quickly identify performance issues, check the activity summary at the top of the profiler when the profile populates. Look for red bars in the activity summary, indicating instances of long tasks during the recording. These red bars help you pinpoint problem areas and focus your investigation.
Lighthouse’s timespans mode is the less intimidating alternative to the DevTools performance profiler. Here’s how to use it:
You will be presented with a list of failed and passed audits. When you click on them, a drop-down menu will appear, and you can see a breakdown of time spent during interaction divided by input delay, processing time, and presentation delay.
Source: Google
Now that you know what to work on, it’s time to roll up your sleeves and begin optimizing.
To guarantee your site a good INP score, you need to ensure that each interaction event runs as fast as possible. Here’s how to achieve it:
1. Avoid recurring timers that overwork the main thread
setTimeout and setInterval are commonly used JavaScript timer functions that can contribute to input delay.
setTimeout schedules a callback to run after a specified time, and while it can help avoid long tasks, it depends on when the timeout occurs and if it coincides with user interactions.
setInterval, on the other hand, schedules a callback to run repeatedly at a specified interval. Because of that, it is more likely to interfere with user interactions. Its recurring nature increases input delay and can affect the responsiveness of interactions.
If you have control over the timers in your code, evaluate their necessity and reduce their workload as much as possible.
2. Avoid long tasks
As you already know, long tasks block the main thread, preventing the browser from executing the interaction events.
To enhance your site’s responsiveness, it is important to minimize the workload on the main thread and consider breaking up long tasks.
By breaking up long tasks into smaller chunks, the main thread gets an opportunity to handle other tasks and respond to user interactions more quickly.
Additionally, breaking up long tasks helps avoid the "jank" effect, where animations and transitions on the page become choppy or stutter due to the overwhelmed main thread. By ensuring that each task completes within a shorter timeframe, the page can maintain a smoother visual experience for the user.
3. Avoid interaction overlap
Interaction overlap means that after a visitor interacted with one element, they make another interaction with the page before the initial interaction has had a chance to render the next frame.
For instance, this can happen when users are typing in form fields, leading to numerous keyboard interactions within a brief timeframe. You can optimize the process by:
1. Consider removing the unnecessary callback
Is the expensive event callback truly necessary? If not, consider removing the code entirely, or if that's not possible, delay its execution until a more suitable time.
2. Defer non-rendering work
Long tasks can be broken up by yielding to the main thread. When you yield to the main thread, you essentially pause the current task and split the remaining work into a separate task. This allows the renderer to handle updates to the user interface that were processed earlier in the event callback. By yielding, you enable the renderer to execute pending changes and ensure a smooth and timely user interface update.
1. Reduce DOM size
A large DOM size is a surefire way to fail the INP assessment. So to ensure that won’t happen, you need to reduce your DOM size, or to put it another way – you need to reduce DOM depth.
Aim for a DOM depth of no more than 1,400 nodes.
You can achieve it by incorporating the following techniques:
2. Avoid excessive or unnecessary work in requestAnimationFrame callbacks
The requestAnimationFrame method tells the browser that you wish to perform an animation and requests that the browser calls a specified function to update an animation right before the next repaint.
The requestAnimationFrame() callback is part of the rendering phase in the event loop and needs to finish before the next frame can be displayed. If you're utilizing requestAnimationFrame() for tasks unrelated to user interface changes, it's essential to recognize that you might be causing a delay in the rendering of the subsequent frame.
So avoid using them when unnecessary.
3. Defer
The ResizeObserver interface reports changes to the dimensions of an Element's content or border box or the bounding box of an SVGElement.
ResizeObserver callbacks run before rendering and can potentially postpone the presentation of the next frame if they involve resource-intensive tasks. Similar to event callbacks, it is advisable to defer any unnecessary logic not required for the immediate upcoming frame.
Based on all the tests we’ve been running in the last couple of months and all the documentation that Google published on INP, we could say that it strongly resembles Largest Contentful Paint (LCP).
A multi-layered Core Web Vital that has a lot of moving parts.
So, since Google first announced the new responsiveness metrics, we started testing and working on features that will improve our clients’ INP scores.
And we’ve been seeing some promising results:
With NitroPack, our clients experience an average 36% improvement in INP.
And that happened all on autopilot. Just by installing NitroPack and thanks to optimization features like:
You can also boost your INP and Core Web Vitals scores without writing a single line of code. Install NitroPack for free and experience the improvements for yourself.
Niko has 5+ years of experience turning those “it’s too technical for me” topics into “I can’t believe I get it” content pieces. He specializes in dissecting nuanced topics like Core Web Vitals, web performance metrics, and site speed optimization techniques. When he’s taking a breather from researching his next content piece, you’ll find him deep into the latest performance news.