Does Hosting Affect User Experience? A Study of eCommerce Performance
The data used in this post was sourced from Yottaa’s Speed of the eCommerce Web Report, a comprehensive report on sites in the Internet Retailer 500 list.
Hosting provider is a crucial component of any web business. Aside from the fact that hosting service carries the weight of being a single point of failure for a website, it’s also something even the most non-technical executive can grasp – and inquire about. When sites go down it’s often the hoster who gets the reflexive blame.
So we know it’s a crucial piece of the website puzzle, but how does the choice of hosting provider impact performance and engagement, if at all? And further, how does hosting affect the performance of businesses in the IR 500?
Hosting Provider and End-User Experience
The chart below shows the seven hosts with the most clients in the IR 500, as well as data for sites that host in-house or declined to name their solution. T2I stands for Time to Interact, a measure of page load time that includes the the full load of the page, to the moment the user can interact with it. In this case, we will consider T2I to be the measure of end-user experience. For perspective, we also added “backend” metrics: Time to First Byte and Time to Last Byte.
Overall these T2I averages run from middling to mediocre. The best group average, sites on Fry hosting, clocks in at 7 seconds, roughly the average for all eCommerce sites, and slower than 57% of sites on the web overall. There is a measurable difference between averages from one host to the next, but it’s not dramatic. If you were to take away the one outlier, Yahoo, the range is 1.2 seconds.
Backend Metrics and Hosting Provider
Now let’s take a closer look at averages for First Byte and Last Byte time, as they are a bit too small to see in the chart above. These metrics are more directly impacted by the host, since they are mostly determined by the time it takes for the server to establish a connection and respond to a request.
Here it seems that any potential correlation between backend metrics and end-user performance breaks down. With a couple of exceptions, the results do not follow the pattern seen in the chart of end-user experience.
But let’s take a step back: these are all exceptional numbers. For some perspective, a 400 MSEC reading for time to Last Byte gets you into the top 16% of all sites on the web for that metric. A 300 MSEC reading means top 3%. Pretty much anywhere in the 200s: top 1%. And keep in mind the figures on this chart are averages of many sites, which makes those high percentile readings that much more impressive.
This leads us to the main observation here: the results for end-user experience metrics and backend metrics are a bad match. Sites on these hosts typically achieve backend performance in the top 3% of all websites, but their average end-user experience isn’t even in the top 50%. If we know these sites all have an inherent advantage because of spectacular backend performance, the question becomes: what goes wrong between the host and the end-user?
We’re not going to put a stake in the ground as to a single factor, but these charts should help:
Neither chart is a perfect match with the chart of user experience metrics, but it trends in the right direction – much more so, at least, than the chart of backend metrics. The weight of assets and weight of images on the site says nothing about the hosting service itself, but instead speaks to the tendencies of the sites that use them. That leads us to our two main conclusions:
- There is no obvious, broadly applicable link between hosting performance and end-user performance
- Factors relating to the content on the site correlate better with end-user experience than other factors
That seems obvious?
If you’re well-versed in web performance then these conclusions won’t be news to you. But we’ve found it’s still very common for business and website owners to regard hosting providers as one of the keys to, and perhaps the most important factor in, achieving good performance. For example, in our study of the performance of sites on various CMS platforms, a common reaction among readers was (to paraphrase): “of course WordPress sites are slow, many users of that platform have cheap, shared hosting solutions, and that’s what makes their sites slow.”
As we did then, we counter with performance data suggesting that the host has far less to do with performance than the tendencies of site owners to load up their sites with heavy content. This study of the IR 500 shows that even sites with phenomenal hosting aren’t always fast ? in fact, they might even be worse than average.
Site owners should be aware that the choices they make about the content on their site have a huge effect on their site’s load time and engagement. If your site has performance problems, audit your content using a tool like websitetest.com. Then work with your marketing team to pare down content and work with IT to streamline content through FEO techniques. (Or, check out Yottaa 4D Optimization Technology for an automated solution.)