PageSpeed Insights (PSI) is simply a instrumentality provided by Google that analyzes a page’s show and provides suggestions for improving its velocity and idiosyncratic experience.

PageSpeed Insights works by analyzing a webpage’s HTML, CSS, fonts, and JavaScript, and provides suggestions to optimize the show of the page. This includes things similar compressing images, minifying code, and reducing the fig of HTTP requests made by the page.

Let’s look astatine PageSpeed Insights much closely.

To start, spell to PageSpeed Insights. Enter a URL and click “Analyze.”

You’ll person the enactment to power betwixt the Desktop and Mobile analysis. Mobile scores volition usually beryllium worse than Desktop scores. Mobile information whitethorn uncover much issues for you to resolve, and that’s what I urge looking at.

Field data

The adjacent conception contains information from existent users of your website. PageSpeed Insights is pulling this from the Chrome User Experience Report (CrUX), which contains the information of Chrome users who opted successful to sharing that data. 

At the apical is simply a tab to power betwixt leafage and root (similar to domain) level data, which aggregates the information for galore pages. You whitethorn not person information for each pages oregon adjacent root data. It depends connected however galore radical sojourn your website and opt successful to sharing this information.

As of April 2023, determination are ~29.5 cardinal origins successful the CrUX dataset.

The adjacent conception is each astir Core Web Vitals (CWV), including a pass/fail assessment. The main metrics are Largest Contentful Paint (LCP), First Input Delay (FID), and Cumulative Layout Shift (CLS). These CWV metrics are the ones Google uses successful its rankings.

The numbers are colour coded to amusement you that greenish = good, orangish = needs improvement, and reddish = poor. In all, 75% of idiosyncratic experiences request to conscionable the threshold acceptable for a fixed metric for it to beryllium considered “good.” These are the thresholds:

If you click “Expand view,” you’ll spot the organisation for each metric.

There are further metrics from the CrUX database that are not presently utilized successful the rankings. These see First Contentful Paint (FCP), Interaction to Next Paint (INP), and Time to First Byte (TTFB). INP volition beryllium replacing FID arsenic a CWV metric successful March 2024.

The numbers are colour coded to amusement you that greenish = good, orangish = needs improvement, and reddish = poor. 75% of idiosyncratic experiences request to conscionable the threshold acceptable for a fixed metric for it to beryllium considered “good.” These are the thresholds:

If you click “Expand view,” you’ll spot the organisation for each metric.

The past conception tells you a spot astir wherever this information comes from. The information is from existent idiosyncratic experiences and is simply a rolling mean implicit a 28-day period.

Lab data

Lighthouse is an open-source instrumentality for measuring the show and prime of webpages. It tin beryllium tally successful your ain browser. But successful the lawsuit of PageSpeed Insights, it runs connected Google’s servers.

You’ll spot respective numbers for Performance, Accessibility, Best Practices, and SEO. All of these truly conscionable cheque for champion practices, but they don’t archer you however good you’re doing successful each of the areas.

Once again, the metrics are colour coded to rapidly springiness you an thought of what is bully and what you whitethorn request to work on. 

For the intent of this article, we’re going to absorption connected the “Performance” section, since that is what SEOs usage the instrumentality for. First up, you person a show people and a screenshot of the page.

You’ll beryllium scored betwixt 0 and 100. The existent people thresholds are:

• Good: scores of 90–100
• Needs improvement: scores of 50–89
• Poor: scores of 0–49

As I mentioned, you tin person a bully people but inactive person a dilatory leafage that doesn’t walk CWV. Other factors, specified arsenic web conditions, server load, caching, and the idiosyncratic device, besides impact leafage load time. 

Only 2% of tested pages people 100. A people of 50 puts you successful the apical 25% of tested pages. 

The people and metrics tin alteration each clip you tally a test. This tin hap due to the fact that of web conditions, load, oregon browsers that marque antithetic decisions successful the page-loading process. I urge moving 3 to 5 tests and averaging the results. 

The people is based connected a calculation and involves weighting respective of the metrics. The weights alteration betwixt Mobile and Desktop. These are the existent scores for Mobile, but cheque the calculator for the latest info.

There’s different metric section, this clip for the laboratory trial metrics. You’ll find LCP and CLS present but won’t find the FID oregon INP metrics from CWV. Those necessitate clicks connected the page, which laboratory investigating doesn’t reproduce. Instead, you tin usage Total Blocking Time (TBT) arsenic a proxy metric to enactment connected improving.

You tin besides click the “Expand view” fastener to get an expanded presumption that includes definitions for the metrics and links with much details.

The past conception tells you a spot astir wherever this information comes from.

If you hover implicit this information, you’ll get adjacent much info connected the trial conditions. While PageSpeed Insights has traditionally utilized a Moto G4 arsenic the trial instrumentality for galore years, it looks similar that is present a Moto G Power. You tin besides get information connected the determination of the test, which volition beryllium North America, Europe, or Asia.

There are immoderate snapshots that visually amusement you however a leafage loaded over time.

If you click “View Treemap,” you tin find the largest files and however overmuch of the codification is unused connected the page.

By default, you’ll spot issues related to each of the metrics. There are buttons wherever you tin filter to issues impacting circumstantial metrics that you whitethorn privation to improve.

The “Opportunities” and “Diagnostics” sections volition amusement you issues that tin assistance your leafage performance.

The estimated savings and improvements they amusement aren’t realistic. There tin beryllium different blockers truthful you won’t get the improvements shown or, successful immoderate cases, immoderate betterment astatine each if you hole an issue. Sometimes, you person to hole aggregate issues to really spot an improvement.

You tin click to grow immoderate of the elements. You’ll person immoderate guidance connected however to hole each issue. The recommendations tin alteration based connected the strategy that’s being tested. For instance, I tested a leafage connected our WordPress blog, and I saw WordPress-specific guidance.

The tips are utile for translating immoderate of the issues into presumption you whitethorn person heard. For instance, the “defer offscreen elements” contented tells you that you should beryllium lazy loading images. You tin past hunt for a plugin successful WordPress that handles lazy loading.

There is further accusation that shows what the LCP representation is, what elements are causing CLS, and what elements are blocking the main thread (what you request to trim to amended FID/INP). This accusation tin assistance you people fixes toward these elements.

There’s besides a conception for passed audits, showing you wherever you’re already doing a bully job. You whitethorn inactive beryllium capable to amended these. But you’re apt amended disconnected spending clip connected different issues.

PageSpeed Insights has a large API. It lets you propulsion the tract information from CrUX and the laboratory information from the Lighthouse test. You tin besides get page-level CWV information successful bulk that tin lone beryllium accessed via PageSpeed Insights.

The occupation is that not everyone has the skills to query the information successful bulk and store it. But we marque it casual for you successful Ahrefs’ Site Audit. Follow the instructions for mounting up CWV successful the Crawl settings.

When you tally a crawl, we’ll propulsion the information from PageSpeed Insights into the Performance report, and you tin drill into problematic pages.

We besides amusement the information from erstwhile crawls, which you tin usage to show show implicit time. Additionally, you tin spot the organisation of affected pages for each idiosyncratic metric.

For valuation and monitoring, I’d usage Ahrefs, arsenic shown above, to spot the problematic pages and trends, arsenic good arsenic the Core Web Vitals study successful Google Search Console (GSC). 

The payment of GSC is that it buckets akin URLs. For the bucketed pages, you volition apt beryllium moving successful 1 strategy oregon template. And erstwhile you hole the issues, you volition hole them for each of the pages successful that bucket.

Once you cognize what pages oregon templates you request to enactment on, you tin usage the guidance from PageSpeed Insights to marque improvements. I’d besides urge checking retired our guides connected LCP, FID, and CLS for however to hole assorted issues.

To cheque the improvements, you tin usage PageSpeed Insights oregon tally different crawl successful Site Audit to get the information successful bulk. The CWV information volition instrumentality longer to amusement the interaction of immoderate changes due to the fact that it is simply a 28-day average, truthful usage PSI oregon the PSI information successful Ahrefs to cheque if the changes you made improved the laboratory trial metrics.

If you person immoderate questions, connection maine on Twitter.