In order to measure loading times or analyse web performance, many of us turn to PageSpeed Insights. But have you ever wondered how the scores are calculated? What does it really mean to score 50, 70 or 90? And is it even possible to reach 100? Let’s take a look at how Google does its maths.
Lighthouse, a light in the darkness
Having recently explored the workings of Test My Site, it’s time we looked more closely at PageSpeed Insights.
These are the web performance indicators collected by Lighthouse and factored into your score on PageSpeed Insights, listed by weighting (the rest is explained here):
It’s immediately apparent that TTI is weighted heavily, despite being a relatively recent indicator that has yet to be set in stone. As such, even though we are all seeking to improve our scores, this provides another reason to adopt a broader perspective.
As a reminder, this is what each metric represents:
When testing a page’s URL, PageSpeed Insights will first produce scores between 0 and 100 for mobile and desktop. A pair of tabs in the top left of the results page allow you to switch between them, displaying the mobile score by default.
So, what do the scores actually mean? It’s simple: a score from 1 to 49 is considered slow, 50 to 89 is considered average, and 90 to 100 is considered fast (if the score is 0, chances are Lighthouse encountered a bug).
The rating scale is calibrated according to measurements collected on the world’s largest sites by HTTP Archive. The highest score possible is 100, which represents the 98th percentile; a score of 50 represents the 75th percentile. In other words, a score of 50 still ranks your website among the top 25% in terms of performance. Nevertheless, the orange colour code can give the impression that this is an average or poor result.
The highest scores are rarely attributed due to the scoring methodology employed, which always assumes the worst-case scenario. Specifically, the mobile score provided by default is often lower than the desktop score, because mobile web performance is harder to optimise. After all, there are greater network constraints on mobile, and a smartphone has less processing power than a desktop computer. In addition, Lighthouse actually simulates a 4G connection that is considerably slower than much of the network in France.
Testing the same page several times yields different scores. Why?
Have you ever tested the same page multiple times with different results? As explained in Google’s Scoring Guide, your Lighthouse/PageSpeed Insights score may vary because conditions themselves may differ between tests: network quality, third-party scripts, ads, etc. Lighthouse is also limited insofar as it only takes one-off measurements, whereas genuine statistical data requires repeated measurements to account for potential aberrations.
Harsh but fair
Getting an average or poor score does not mean that the site is unusable – far from it! To illustrate, take a look at the top 20 sites in the mobile web performance rankings for June 2019 (this is a ranking of the most visited French websites)– with home-page scores on PageSpeed Insights in the right-hand column.
Among the most-visited French mobile websites, these are the top 20 in terms of web performance, even though 70% have an average score, only 3 are above 90 and some are even below 50. As we have seen, these scores are variable from one test to the next. While those listed in the table below represent a snapshot at time T, they still provide a reasonable indication of web performance. If you would like to obtain a more reliable score, it’s best to take multiple tests and calculate the median of scores provided – a task that is easily automated:
|Rankings||Website||Webperf Score||Speed Index||PSI Score|
|3||Ouest France||2136||2603|| |
|5||Le Monde||2231||2147|| |
|15||Météo France||2780||3009|| |
|17||France Televisions||3135||3061|| |
|18||La Redoute||3555||3905|| |
PageSpeed Insights uses a strict scoring system that encourages web performance optimisation – and that’s a good thing! Although you shouldn’t interpret these results to the letter (or indeed the number), they do invite reflection on how to improve loading times. Any score “in the red” signifies that it’s time to focus on optimising the performance of your website – and that’s exactly our core business.
In part two, PageSpeed Insights: how to interpret and apply results 2/2, we shall discuss the more specific indicators on PageSpeed Insights, and what to make of the recommendations provided in each section of the results page.
Would like to find out more about web performance indicators
and gain more of an insight into the key metrics?