Page speed against rivals: reading PageSpeed Insights calmly

Two kinds of data

PageSpeed Insights reports two different things, and mixing them up causes most of the confusion:

• Lab data comes from Lighthouse running in a simulated environment. It is useful for debugging because it is repeatable and shows what is slowing a page down.
• Field data comes from real Chrome users, via the Chrome UX Report, over the previous 28-day collection period. It covers Largest Contentful Paint (LCP), Interaction to Next Paint (INP), Cumulative Layout Shift (CLS) and First Contentful Paint (FCP).

Both are reported for mobile and desktop. Field data tells you what visitors actually experienced. Lab data tells you why. Smaller sites may not have enough traffic for field data, in which case you only see the lab view.

Compare the same kinds of page

A homepage against a homepage is fair. Your heavy product page against a rival's light category page is not. Pick matching pages for you and three rivals:

• The homepage.
• A main category or service page.
• A typical product page, or for service firms, a typical service page.

Test each on mobile first, since that is where most visitors arrive and where differences are usually biggest. Record the field data where it exists and the lab score separately.

Lab scores move around

Run the lab test several times and you will often see the score change. Network conditions, server load and third-party scripts all vary. Do not compare one run of yours against one run of a rival's and draw conclusions from a single pair of numbers. Run each page a few times and look at the range.

Field data is steadier because it is collected across many real visits rather than one simulated load. If field data exists for both you and a rival, it is the better basis for comparison.

What a gap does and does not tell you

If a rival's product pages pass the field data checks and yours do not, that is a useful observation. It is not proof that the rival sells more because of speed. Speed is one factor among many, alongside price, range, trust and how clear the page is.

Hypothetical example: a homeware shop finds its product pages have a much slower LCP than three rivals'. The lab report points to large, uncompressed hero images. Compressing them is a low-effort change with a plausible upside, and it can be done without touching the design. That is the kind of gap worth acting on first.

By contrast, a small difference in a lab score between two fast sites is rarely worth chasing.

Usual suspects

When the lab report points to problems, these are the most common causes on shop and service sites:

• Large images that are not compressed or not sized for mobile.
• Too many third-party scripts: chat widgets, review widgets, tracking tags and apps that were installed and forgotten.
• Fonts loaded from several sources.
• Layout that jumps as banners or images load, which shows up as CLS.

Most of these can be fixed without a rebuild. Our post on rebuild or fix covers when the answer really is a new site.

Record the comparison so you can repeat it

A single test is a snapshot. The value comes from repeating the same comparison later and seeing whether the gap has closed. Keep a simple sheet with one row per page and columns for:

• The site and the exact page address.
• Mobile field data for LCP, INP and CLS, where it exists.
• The range of lab scores across your runs.
• The top issues the lab report listed.
• The date you tested.

Test the same pages each time. If a rival redesigns, note it, because a jump in their numbers may be down to the new site rather than anything you did.

For service firms with smaller sites, field data is often missing for everyone. In that case the lab report is still useful, but treat it as a list of things to look at rather than a league table. A solicitor's or clinic's site that loads its main content quickly on a phone, without the layout jumping around, has done most of what matters.

Finally, keep speed in proportion. A fast page with unclear copy and no reviews will still struggle. Fix the obvious speed problems, then put your effort where the comparison says you are furthest behind.

Next step

An Intel loop includes speed and basic SEO as one of its eight scored categories, alongside mobile experience, with quoted evidence for each score, so you can see speed in the context of everything else. For a free starting point, run the free Insight Report.