The median load time for home pages of the top 500 US retail websites (as ranked by Alexa) keeps slowing down, clocked at 8.56 seconds during the Fall, up from 7.48 seconds in the Summer and from 7.36 seconds in the Fall of last year, per results [download page] from Radware’s latest E-Commerce Page Speed and Web Performance study. In fact, 38% of the Alexa Retail 500 took at least 10 seconds to load.
The slowdowns are continuing as pages grow in size, with the median page about 15% bigger in the Fall than in the Summer. Specifically, the median page was 1,258 kb in size and contained 92 resources, up from 1095 kb and 89 resources.
Median Time to Interact (TTI), a metric introduced in the previous report that analyzes the “point at which a page displays its primary interactive content,” has also slowed. The metric is designed to be a key gauge of how quickly a page can deliver a satisfactory user experience, according to the researchers, as well as how rapidly the page can accomplish the site owner’s goals. Among the top 100 e-commerce sites, the median TTI stood at 5.3 seconds in the Fall, up from 4.9 seconds in the Summer. What’s more, 26% of the sites took at least 8 seconds to become interactive, a marked rise from 9% in the previous report.
The researchers identify 3 design best practices that are affecting usability:
- Feature banners being the last page elements to load;
- Feature banners loading last with the call-to-action at the bottom, making it the final element to render; and
- Feature banners loading last, without a call-to-action.
- Some top retail websites continue to not employ best practices, although there has been significant improvement. All of the top retailers now enable keep-alives (up from 90% in the Summer), while 79% use a content delivery network (CDN), up from 58.4%. Meanwhile, 45% of the top 100 sites failed to compress images, and only 6% used progressive JPEGs.
- Performance is slowing across each of the 3 major browsers, and is now at 7.91 seconds for Firefox 22 and Internet Explorer 10, and 8.56 seconds for Chrome 29.
About the Data: The tests in the study were conducted using an online tool called WebPagetest ”“ an open-source project primarily developed and supported by Google ”“ which simulates page load times from a real user’s perspective using real browsers. Radware tested the home page of every site in the Alexa Retail 500 nine consecutive times. (The system clears the cache between tests.) The median test result for each home page was recorded and used in the calculations.
The tests were conducted on August 29-30, 2013, via the WebPagetest.org server in Dulles, VA, using the following browsers on a DSL connection:
- Internet Explorer 10
- Firefox 22
- Chrome 29
In very few cases, WebPagetest rendered a blank page or an error in which none of the page rendered. These instances were represented as null in the test appendix. Also, in very few cases, WebPagetest.org rendered a page in more then 60 seconds (the default timeout for webpagetest.org). In these cases, 60 seconds was used for the result instead of null.
To identify the time to interact (TTI) for each page, Radware generated a timed filmstrip view of the median page load for each site in the Alexa Retail 100. TTI is defined as the moment that the featured page content and primary call-to-action button or menu was rendered in the frame.