All online retailers strive to provide a great experience for both online and mobile users. Not all do.  In fact, within the Internet Retailer 500, a list of the largest U.S. online retailers in terms of annual revenue, the average desktop page load time on a cable or Fios connection is a whopping 8.6 seconds.  That’s leagues away from the general web benchmark of 2.0 seconds, a page load time that begets great user experiences.

While some of these 500 retailers are successful in spite of poor performing applications, others are role models for any online business that is seeking to get ahead by providing a great end user experience.  In order to find out what makes these top apps tick, we compared the top quintile of the 500 (in terms of page load time) with the rest of the bunch, across a broad series of metrics.

We will explore each of these categories in detail in the coming weeks in individual posts – this is an overview.  Let us know in the comments if you have any questions you’d like to see covered, and we will work them in.

To set the stage, here’s how much better the top quintile perform on desktop than their counterparts:

 As you can see, the top 20 percent of sites perform more than 50% better.  What traits do they have that get them there? Let’s see.

1. Lighter, less complex apps

The top performing IR 500 apps are significantly lighter and less complex than the rest.  Like all successful eCommerce sites, they provide an experience that enables the user to navigate, research, find items, and complete a purchase – yet they are able to provide that experience with forty percent fewer overall assets, and nearly half the overall weight (kb of data).

It has long been known that weight and complexity correlate with performance.  It makes sense intuitively, as well as being borne out in broad studies of thousands of websites. This data shows that the IR 500 are no different – there’s no magical silver bullet that has enabled them to have their cake and eat it too.  Just like the rest of the web, lighter and simpler means better performing.

2. And same goes for mobile

Given that end-user experiences on mobile devices are continually below expectations, this point bears exploration in the mobile context.

For mobile sites, we looked again at the best 20% of samples.  The graphs look similar to those on the desktop but with lower numbers overall for asset counts. This is likely a result of the presence of customized “m.dot” sites that tend to have smaller footprints than their desktop counterparts. And in the case of responsive sites, there are sometimes server-side optimizations that help to reign in the amount of content sent to mobile devices to improve performance.  Overall, the top sites are lighter and less complex.

Speaking of the m.dot/responsive divide, the top quintile for mobile performance are actually not much different than the rest for frequency of one or the other.  64% of the top quintile employ responsive design (or perhaps no mobile solution at all). The figure is 69% for the rest.  We’ll look further into the performance differences between responsive and m.dot sites in an upcoming post.

3. User experience technology

This is actually an anti-trait. Delivery is an important component of performance, and a large expense for most major retailers.  But how does it affect the distribution of performance? The truth is, not much.  Here are the most common CDN providers listed by IR 500 companies, including all with more than 5 different companies as customers.  By far the largest categories are Akamai, In-house, and NA (meaning the company either does not use a CDN or does not list it publicly).

It?s undeniable that CDNs are important for scalability and maintaining performance for end users in all regions of the world. But our data show that for the top quintile sites it doesn’t matter which CDN they use. And for the rest the difference isn’t immense.

In a later post we’ll explore other implications for supporting technologies, such as hosting and eCommerce platform.

Follow our blog for more!

In the next few weeks look out for new posts exploring this data, in an effort to find out what traits are common among great performing apps. And again, let us know in the comments if you have any questions you’d like to see covered!