Traditionally, when we look at web performance, we create a time-series graph that focuses on dimensions like browser or page template. These are all technically-oriented measurements that are collected automatically based on what is readily available. Last year at the Velocity conference, I met Anh-Tuan Gai from WebPerf IO who showed me a more business-oriented approach to visualizing performance data. I found his approach very interesting and asked him to collaborate on sharing the approach for Web Performance Today.
At WebPerf IO, they start the performance measurement activity by segmenting users into groups based on business value. For example, if the site is particularly interested in generating traffic from sources like Facebook or search engines, then they segment users into these groups prior to looking at the performance data. To help illustrate this, Anh-Tuan provided anonymized data.
Here is a look at the data from a traditional time series perspective:
This graph shows a time series of high traffic pages from a sample site. There are three different page templates (home, section, and article) and the graph shows the performance of each over a nine week period. While informative, notice that it is difficult to understand how page performance impacts the business by looking at just this graph alone.
In an attempt to uncover how a business is impacted by speed, it is often common to introduce bounce rate to a visualization. A low bounce rate percentage normally indicates higher user engagement – users stay on the site longer and view more pages. A high bounce rate indicates that users are abandoning the site quickly and are not as highly engaged.
The above graph shows a traditional view of bounce rate in % by load time (in seconds). We have also included the traffic volume (page view count) to illustrate that most of the users experience a page that loads in ~2.5 seconds with a bounce rate of ~24%.
It is becoming more evident that bounce rate climbs as pages takes longer to load. However, this is still a relatively generalized graph, where it is hard to see any evidence of how our users are reacting to performance.
Now for some excitement: What if we segment the users into the business groups that our marketing efforts are targeting? In this case, we’re looking at users who are coming from Facebook and search results.
The first group – ‘direct’ (light blue line) – are users who are navigating directly from a bookmark or entering the url in the address bar. The second group – ‘search’ (darker blue line) – are those navigating to the target site from a search engine like Google. The third group (orange line) is the most interesting, made up of users that navigate to the target site from Facebook.
In this case, we can see that users from Facebook, a segment our business is particularly interested in, are much more sensitive to performance. For the first two groups, the bounce rate for pages that load in 4 seconds is in the 20% – 30% range, but more than 60% of users who navigated from Facebook have bounced at this point. Interesting, and not at all apparent from the traditional views.
After talking more with Anh-Tuan, his experience is that often the segments we care about most are the most sensitive to performance. Considering this, many businesses looking at performance metric may be underestimating the impact of performance on their most important users.
For more on the importance of speed and how the context of speed can affect very real business goals, especially in ecommerce, I invite you download our most recent State of the Union: Ecommerce Page Speed and Web Performance Report. The report offers data on the most popular online retail sites stack up against each other. There are also some great tips you can implement to help take charge of your site’s performance.
The post The (Context) Sensitivity of Speed appeared first on Web Performance Today.