A site that loads quickly on your computer can easily frustrate half of your mobile visitors. Evaluating a website’s performance is not just about running a quick speed test: it requires looking at multiple angles, from technical responsiveness to accessibility, to get a truly useful diagnosis. Here’s how to structure this approach without drowning in data.
Interaction to Next Paint: the performance metric that classic tests ignore
Have you ever clicked a button and waited a second or two before the page responded? This delay now has a technical name: Interaction to Next Paint (INP). Google has integrated it into the Core Web Vitals to measure actual responsiveness after loading.
Most guides focus on LCP (Largest Contentful Paint), which assesses the speed of displaying the main content. INP complements this picture by capturing what happens when the user acts: clicking, filling out a form, pressing a button. A site can display its page quickly while taking a noticeable amount of time to respond to interactions.
To identify these bottlenecks, there are several methods for evaluating your website that take INP into account alongside other indicators. The third pillar of the Core Web Vitals, CLS (Cumulative Layout Shift), measures visual stability: those annoying shifts when an element moves under your finger just as you’re about to click.
A good performance audit crosses these three metrics rather than relying on a single score.
Lab data and field data: why a high score can be misleading
Launching PageSpeed Insights and getting a green score gives a false sense of security. This test simulates loading under controlled conditions, with a standardized connection and device. Your actual visitors, however, are browsing from various smartphones, on more or less stable mobile networks.
The Core Web Vitals are assessed from field data, meaning the experience actually lived by your users. A site can therefore pass a one-time lab test while remaining slow or unresponsive for a significant portion of its audience.
How to cross-reference the two types of data
Synthetic testing tools (like a manually launched audit) are useful for diagnosing a specific problem: an oversized image, a blocking script, a poorly optimized font. They allow for reproducing and isolating a technical defect.
Field data, accessible via Google Search Console or the Chrome User Experience report, show what your visitors actually experience over several weeks. It is this second source that determines how Google evaluates your page.
- Use lab tests to identify the technical causes of a slowdown (scripts, images, blocking resources).
- Consult field data to spot pages where user experience degrades on mobile or on slow connections.
- Compare the two: if the lab score is good but the field data shows a high INP, the problem likely comes from a script that triggers after the initial load.
This approach avoids fixing problems that only exist in a testing environment.
Digital accessibility: a regulatory evaluation criterion since 2025
Evaluating a website by limiting it to speed and SEO means missing out on a whole aspect of performance. Since June 28, 2025, the European Accessibility Act applies in France to several services aimed at consumers: e-commerce, banking services, telecommunications, transport.
In practical terms, the companies concerned must make their journeys usable for people with disabilities, inform users of the accessibility level of their site, and provide a way to report a blockage. Micro-enterprises providing services benefit from an exemption under certain conditions.
Integrating accessibility into the performance audit
Accessibility is not a separate project. A contact form that is unusable with a keyboard, insufficient color contrast, or navigation that is incomprehensible without a mouse penalizes both user experience and regulatory compliance.
- Ensure that each page offers functional keyboard navigation, without traps (an element that captures focus with no way to escape).
- Test color contrasts between text and background, especially on action buttons and links.
- Make sure that meaningful images have descriptive alternative text.
- Check that forms display clear error messages associated with the relevant field.
A fast but inaccessible site excludes part of its audience and is exposed to compliance obligations.
Building a website audit that leads to concrete decisions
Accumulating reports is pointless if no one acts on them. The classic trap: launching five different analysis tools, gathering dozens of recommendations, and then prioritizing nothing.
Start with the pages that receive the most traffic or those that carry your conversion goals (contact page, product page, quote request). Focus the analysis on these pages before moving on to the next ones.
For each page, ask three simple questions. Does the main content load quickly (LCP)? Does the page respond without delay to interactions (INP)? Do elements remain stable during loading (CLS)? If any of these answers are negative in the field data, you have your priority.
Prioritize corrections by impact on the user journey, not by technical ease. An oversized image on your homepage weighs more than an unoptimized script on a little-visited secondary page.
Evaluating a website should become a regular reflex rather than a one-off annual audit. Performance fluctuates with every content update, every feature addition, every theme change. Monthly monitoring of Core Web Vitals on your strategic pages is enough to detect degradation before it affects your SEO or conversions.
