Google Page Speed Insight

Now that we understand the methodology behind the PageSpeed Insights score and what this means, we ought to examine the additional results to determine the relevance of the recommendations under Opportunities and Diagnostics – and whether these should be applied.

Breaking down the PageSpeed Insights results page 

FIELD DATA

Under the Lighthouse score, PageSpeed Insights contains a section called “Field Data”. This consists of CrUX  data collected over 30 days. Keep in mind that this data represents the experience at the lowest end of the spectrum: it is based on the 90th percentile for FCP and the 95th percentile for FID. Respectively, some 90% and 95% of visitors will experience faster times than those reflected by these metrics.

You may well ask why Google does not see fit to include median results. Well, there is a case to be made for providing extreme values to encourage loading time optimisation. Although the data is not the most representative, it does highlight the most critical cases in need of urgent attention. Another helpful feature is the visual display of the value distribution over 3 groups: “fast / average / slow”. In this example, the majorhttp://fasterize.com/fr/blog/les-risques-webperf-focus-sur-javascript-vs-les-mobiles-dentree-de-gamme/ity of users experience fast browsing speeds:

Clicking on Show Origin Summary will display the data for the entire domain name, whereas Field Data only covers the page tested. 

Note that the FCP under Field Data is different from the FCP under Lab Data, which we shall cover in the following section. It’s essentially the same metric applied using a different procedure.

Also bear in mind that, if your site traffic is private and is not part of the CrUX dashboard, Google will not collect real-world data, and you will not have access to the information under this section. In this case, the following message will be displayed:

Lab Data

Lighthouse calculates “laboratory” data by extrapolating results based on a native connection using an algorithm (as opposed to the simulated connection seen with WebPageTest). As such, this constitutes synthetic data, which explains why the results for certain indicators will differ from those in the Field Data section (which is based on RUM data), in addition to the margin of error due to this extrapolation of results.

 

As we discussed in our previous article on PageSpeed Insights, the results are weighted in order to calculate the score of 0 to 100, and ought to be considered from a broad perspective. The same can be said of the advice provided in the following sections of the results page (Opportunities and Diagnostics), which only relates to the page that you tested. Although these recommendations will often be valuable, the conditions for their implementation are not explained, and, above all, they do not provide a systemic overview. Additionally, our tests indicate that the Estimated Savings are highly optimistic – or indeed unrealistic. For example, the recommendations with regard to image compression do not account for perceived quality:

How does Fasterize enhance the results marked in red and improve your overall PageSpeed Insights score?

As you can see in the screenshots above, an orange or red colour code is used to denote avenues for improvement. 

They may include best practices that really ought to be applied, or recommendations that you may be unable to implement because you are not necessarily in control of the relevant factors. 

The same applies to our clients who benefit from automated optimisations via the Fasterize platform, as it is not designed to address 100% of the issues highlighted on PageSpeed Insights.

For instance, if the tool recommends that you optimise third parties, this is unfeasible because the relevant scripts originate from third-party publishers and are highly complicated to optimise through our engine.

Below you will find a summary of everything that Fasterize can improve in terms of the web performance recommendations on PageSpeed Insights. For more information about the relevant operations, please check our Support.

 

Lighthouse Notification Can be fixed with Fasterize
Does not use HTTPS Yes
Does not redirect HTTP traffic to HTTPS Yes
Does not respond with a 200 when offline No
Page load is not fast enough on mobile networks Possibly
Reduce server response times (TTFB) Yes
Reduce JavaScript execution time Partially
Preload key requests Yes (manually)
Preconnect to required origins Yes (manually)
Ensure text remains visible during webfont load Yes
Serve static assets with an efficient cache policy Yes
Avoid enormous network payloads Yes
Defer offscreen images Yes
Eliminate render-blocking resources Partially
Minify CSS Yes
Minify JavaScript Yes
Remove unused CSS No
Serve images in next-gen formats No
Efficiently encode images Partially
Enable text compression Yes
Properly size images Yes (manually)
Use video formats for animated content No
Avoid an excessive DOM size No
Does use HTTP/2 for all of its resources Yes
Does not use passive listeners to improve scrolling performance No
‘robots.txt’ is not valid Yes

In conclusion, there are two important points to remember when using PageSpeed Insights, which is a valuable tool, however...

  • the results should be kept in perspective;
  • when deciding which optimisations to implement – and how – nothing can replace the insight and opinion of a human expert!

To find out how Fasterize can improve your loading times in practice,
simply request a demo:

Ask a demo!

 


Hello SMX Paris !