Here are 2 questions:
- Is the speed of a website important?
- Which KPI should you use to measure it?
While the answer to the first question is usually exactly the same from anyone answering it (‘yes!’), the answer to the second varies.
Universally, almost everyone agrees website speed is important:
Customers also say it’s important (at Satalytics we have feedback from hundreds of thousands of ecommerce customers across many sites, ‘speed’ is a regular complaint where sites aren’t fast. ‘fast’ appears in feedback more frequently where satisfaction scores are positive; ‘slow’ more frequently where scores are low).
But there’s little agreement on which KPIs to measure. Here we’ll suggest a simple, useful metric that’s hidden away in Google Analytics, and explain why the ‘standard’ Google Analytics site speed KPI isn’t so great.
Which Metric?
Here’s data from Google themselves on the effect of poor load speed:
In other words: optimising your site to get all of your page loads under 3 seconds can, in theory, save roughly half of your mobile visitors from abandoning. Therefore, as a simple, actionable KPI, we’d suggest ‘percentage of pages loaded in less than 3 seconds‘ should be something you track and work toward maximising.
But Google Analytics does not make it straightforward to see how many of your visitors see pages loaded in less than 3 seconds.
Here’s what Google Analytics does show you as standard, why it’s not great, and how to get to the ‘percentage of pages loaded in less than 3 seconds’ metric:
The Standard Metric: Average Page Speed
The default data you see in Google Analytics is based around ‘average page speed’. That’s kind of useful, but has several flaws. Here are 3 notes to explain:
- Google Analytics – by default – does not record the speed of every page load. By default, it only tracks speed for 1% of pages viewed. That means, if 100 users load a single page on your site, the speed date you see is for just a single user. Perhaps that user had a superfast connection; or perhaps that user was on a train with extremely poor connectivity. The user’s situation affects your perceived load speed there as much as the technical aspects of your website.
- Google Analytics – by default – uses ‘average’ load time. That means if it records data for 10 pageviews, and 9 of those loaded the page in 2 seconds, and the 10th was a terrible connection that loaded the page in 35 seconds, the page speed number you see is 5.3 seconds.
- The above two issues compound together. They mean that it’s possible to have a page where 1,000 users visit, and just 1% (10) of those pageviews record speed. And for those 10, if one load is particularly slow, it skews the average to such an extent that you are left with the overall impression that all 1,000 users experienced a terribly slow site.
Misleading Data: An Example
This sounds rare, but here’s an example of it in action:
Usually, the page speed of this page (an important page on the site) is 2-3 seconds. For a particular period, it looked like this:
As you can see, the dot there is just under the ‘6’ second mark. So why had the website suddenly jumped from loading in a couple of seconds, to loading in just under 6 seconds?
The answer is: It had not – this is simply a trick of the metrics. Page load speed was very fast for most of the period, but on the 23rd and 25th of December Google Analytics reported it in such a way that it appeared disastrous: ‘Average Page Load Time’ as reported by Google Analytics was 25 seconds on one day, and even higher on another.
The answer is simple: For that particular page, on those days, Google only recorded one or two page speed measurements, and those happened to be from users on extremely slow connections.
In other words: You can drill into the data in Google Analytics to help understand site speed, but relying on the top level ‘Avg Page Load Time’ metric is misleading.
An Answer: The Best Page Speed Stat in Google Analytics
So what should Google Analytics users to do pay attention to site speed if the primary metrics are misleading? There is an answer, but it’s hidden away.
Percentage of Pages Loaded in Less Than 3 Seconds
For users who are interested in page speed, but want a KPI they can track each week or each month to understand whether they’re improving or falling behind, the following path leads you to a well-hidden set of numbers that can help:
Step 1: Go to the ‘BEHAVIOUR’ menu, navigate to ‘Site Speed’, and navigate to ‘Page Timings’.
Step 2: Once there, toward the top left of the report, you’ll see a subtle link with the anonymous-sounding title ‘Distribution’:
Step 3:
Once there, instead of seeing a basic ‘average page load time’, you’ll see the actual distribution of how many pages on your site were loaded in ‘less than 1 second’, ‘1-3 seconds’, ‘3-7’ seconds, and beyond:
THIS is an actual, useful report: data showing the breakdown of how fast your pages loaded for actual visitors. In the case above, we can see that just 0.85% of pages loaded in less than a second, a further 31.4% loaded in less than 3 seconds. In other words, for the site above, ‘pages loaded in less than 3 seconds’ is currently at 32.65%, or just less than a third.
Summary:
- The standard Google Analytics ‘page speed’ KPI is interesting, but misleading.
- Google themselves believe pages loaded in less than 3 seconds give you a big advantage vs slower loads (saving 53% abandonment).
- The ‘percentage of pages loaded in less than 3 seconds‘ KPI is present in Google Analytics, but is tucked away.
- We recommend finding it, tracking it, and working toward optimising your site to push as many page loads into the ‘less than 3 seconds’ buckets as possible.