August 11, 2015
A few years ago, I read a short but good article on the speed of eCommerce sites over at Pingdom. Pingdom had created a small test to benchmark the loading speeds of the top 100 eCommerce websites. At the end, the conclusion was that there was still a lot more to be done to enhance performance.
Like the parrot I am, I have been saying this over and over again; for SEO and CRO load times are becoming increasingly important. We have all read how speed affects conversion, but still, to recap here’s some interesting data:
- A study at the University of Nebraska-Lincoln shows that tolerable waiting time for loading a page is about two seconds.
- This study has come to the conclusion that the demand for superior loading speeds has increased over the last few years. A page clocking in at six seconds would mean a 40 percent hit in conversions four years ago. Today, I guess that figure would be well above 50 percent.
- Nearly half of all Web users tend to abandon a site that doesn’t load in three seconds, and 79 percent of shoppers wouldn’t return to the same site again should they have a poor shopping experience.
A Speed Test of the 100 Biggest Ecommerce Sites
Just like Pingdom, I chose to perform tests on the top 100 shopping sites according to Alexa. The absolute majority of these sites are U.S.-based, which is why I chose to use their Dallas data center for the tests. Since I have been writing a lot about page grading – Google Pagespeed Insights and Pingdom lately, I also chose to check those when doing the tests.
I, to the best of my ability, made sure I was testing the actual index page and not some country selection entry page. Every test was performed three times from which the best loading time was taken.
The checked sites were:
Loading Times for the Top 10 and Bottom 10 Sites
This is by far the most important figure, or rather – it is the figure we are going to look at. Everything else is a factor affecting the time. The average loading time for all 100 sites was 2,630ms. The top 10 sites averaged a 748ms loading time.
Google coming in at the third position is no surprise considering its simple search page. Astonishing results are, however, delivered by Zappos with a mere 326ms of loading time. I had to check this several times and couldn’t exactly figure out how it was accomplished. It, along with Amazon, are the masters of conversion, but the figures are still impressive.
Hm.com is also doing a great job clocking in at 754ms. Other stores in the same segment are not performing as well:
4330ms – Mango
3180ms – Gap
2540ms – Urban Outfitters
1740ms – Nordstrom
The bottom 10 sites had an average loading time of 7,103ms. That’s SEVEN seconds, which comes as no surprise as most of them make hundreds of server requests. You however have to factor in location. Trademe.co.nz, for example, host their site in Wellington, NZ.
Least and Most HTTP Requests
This is, according to me, the factor having the most impact on site performance. All 100 sites had an average of 190 requests, while the top 10 had an average of 55.
Netflix gets to the conversion stage quickly — no distracting stuff — bells and whistles. Wiley gets the point of having a lightweight site, sparingly using images and rather intelligently “designing with HTML.” Hm.com once again shines, getting in at place five with just 63 requests.
The bottom 10 sites had an average of a whopping 442 requests. At first glance, these values are simply ridiculous, but one has to keep in mind, we also have to consider render time. Loading 400 files might seem much, but in that we also have to factor in lazy loading of images and other objects.
Size of the Sites
The average size for all 100 sites was 2,300KB. The top 10 sites had an average size of 620KB. Bottom 10 sites had an average size of 6,980KB. It’s no surprise they all average a 3.5 second loading time…
Google PageSpeed Score vs Site Load Time
The average Google PageSpeed Insights (desktop) score for all 100 sites was 53. The average score for the top 10 fastest loading sites was 77.7. The 10 slowest sites averaged a score of 59.9.
So how does Today’s Data Compare to those of the 2012 Pingdom Tests?
The above cross section of data is interesting, but how does today’s data compare to that of 2012? To check this we took all the results from the Pingdom article and made a new test. This time, like in the original test, choose to use the Swedish datacenter.
Interestingly, many of the figures are worse, pretty much across the line. Three years is a lot of time when it comes to site design and Internet infrastructure. Having a site “weighing” more might not be such a big thing considering how much faster, especially mobile connections have become the last few years. The average number of requests didn’t improve either, rising from 110 to 174. In this case you should, however, factor in the many objects that have been added for the sake of mobile usability. This would, of course, (I think) be so in the case of responsive sites, and not those running adaptive versions.
Just as Pingdom concluded back in 2012 there is a lot of room for improvement. Though, when viewing the figures you have to keep several other factors in mind – factors that are not easily seen, isolated or calculated:
- Data center location.
- Actual rendering time for what is displayed in the viewport. Measuring the total loading time of a site does not always do it justice.
- We don’t know, exactly, what the Pingdom / Googlebot sees. There is a chance of heavier page elements not being served to the bots.
- Mobile versions. Those sites that were responsive would have more objects, and so more requests than those utilizing an adaptive approach.
A lot of sites have request figures shooting through the roof. While I think this is one of the biggest culprits when it comes to slow loading sites, the future might be different. The new HTTP2 standard is coming, and with it, requests will be handled much more efficiently.
Nevertheless. It will be interesting to see what is to come in the approaching years.
Philip Blomsterberg is a Swedish SEO and owner of the newly started company Intripid SEO. Philip has more than 20 years of experience in various online projects, such as eCommerce, affiliate marketing, development of software and, obviously, SEO. He has his own SEO blog and also post on sites such as Majestic, Ahrefs and Smashing Mag.