As website owners, we all know intuitively that our websites need to be fast. That’s not just an intuition, as it has been proven time and again that a fast website is good for business.
According to a test done by Kissmetrics, a 1-second delay in page response can result in a 7% reduction in conversions. Moreover, 79% of users who aren’t pleased by a website’s performance say that they will not return to that site.
The issue is even more pronounced with mobile users who not only have a slower internet connection and weaker computing power, they are also less patient than desktop users.
In addition, website performance is a ranking factor for Google, which favors fast websites in their search results.
Just How Fast Does a Website Have to Be?
OK, so a site needs to be fast; but just how fast? How do we measure its performance?
The usual answer to this question would be to measure the “load time”, but that can sometimes be tricky because there are different approaches to doing so:
We could measure the time until a page is fully loaded with all its assets, images, and scripts – but do we really care if a script is loaded “behind the scenes” or an image is loaded below the fold?
- We could measure the time until just the “above-the-fold” content is loaded – but fold location varies per device…
- We could measure the time it takes for the browser to start receiving its HTML (time till the first byte), but this fails to illustrate how users actually see our website…
In short, there is no single way to measure load time.
An alternative method for measuring load time is to measure a website’s implementation of performance best practices. There are various tools that scan a website and evaluate it against a predefined set of performance best practices. While this approach is good for pinpointing underlying performance problems, some of the tools that utilize this approach give too much importance to things that have no effect on the user experience – and thus can be inaccurate.
Webpals’ Hybrid Approach to Measuring Website Performance
In evaluating website performance here at Webpals, we take a two-sided approach:
- On one hand, we use a page-load measuring tool to get a single result for the entire site’s page load time
- On the other hand, we complement it with another tool that measures performance best-practices
Our studies show that although the two metrics don’t have a 100% correlation, trying to optimize our websites for those two KPIs together can increase performance for almost all users, whether they’re using desktop computers or mobile devices, and whether they’re human or, alternatively, search engine crawlers.
To measure page load, we use a Real User Monitoring (RUM) tool. RUM tools, unlike tools that can ping our site every hour, measure the actual load time (in milliseconds) of each and every user. Using RUM tools, we can also segment the data by device types, geo locations, or gain additional insights on performance by splitting the load time to its states (backend time, network time, and frontend time). To obtain a statistically significant result, for each site we take the median page-load time for a time span of at least one day.
To measure the best-practices element, we use Google’s PageSpeed Insights tool. This tool receives a URL of a webpage, analyzes the page in terms of performance best-practices, and issues performance scores for mobile and desktop. In addition to getting a breakdown of the underlying problems that led to this score, this tool also has some SEO value because it gives us an indication of how Google views our website in terms of performance.
Now we have a general idea of what website performance is. We’ve also outlined how we measure the performance of our 2,000+ websites here at Webpals.
In the next article in this series, we’ll talk about how we were able to speed up the majority of those sites using PageSpeed Insights, examining the errors this tool can display and how exactly you can fix them.