eCommerce Closes in on the Three Second Pageload Target

April 12, 2016 — by Matt Young0

main

Application DeliveryWPO

eCommerce Closes in on the Three Second Pageload Target

April 12, 2016 — by Matt Young0

It may be hard to believe, but ecommerce sites have been around in earnest for a little over 20 years – Amazon and eBay were both founded in 1995 (right as the Netscape and Internet Explorer browsers were debuting), preceded the year before by J.C. Penney. In 1997, Dell became the first company to land $1 million in online sales.

Twenty years later, it’s hard to imagine a world without ecommerce. Virtually anything can be, and is, bought online, to the tune of $1.2 trillion U.S. dollars globally in 2013, and an estimated $1.672 trillion in 2015. It may have started with books and music, but it encompasses pretty much everything at this point.

As a transactional medium, ecommerce websites are generally finely-tuned sales mechanisms, with slick links to products and sophisticated algorithms for suggesting other purchases to keep buyers buying.

But what happens if prospective customers never get to that point?

Web performance is a big part of the picture for customers, and studies have shown that up to 57% of users will quit a website that takes too long to load – whether due to a steady uptick in the number of images, complicated JavaScript, or other plugins, as we again found in our research for our latest report, 2016 State of the Union: Multi-Industry Web Performance (Desktop Edition).

We’ve been detailing the web performance of the top ecommerce sites since 2010 in our State of the Union reports, some of which you can find here. In that time, we’ve noticed the impact of page bloat and spiraling complexity on load times and the resultant Time to Interact (TTI) has taken a hit, meaning central images and content aren’t accessible to users right away. This was again evident as we analyzed the top 50 global ecommerce sites, as ranked by information technology company SimilarWeb.

It has an impact: assuming a conversion rate of 2% and an average spend of $115 per person, as the Centre for Retail Research estimates, a site that gets 100,000 visitors a day would lose over $130,000 a day if 57% of its visitors bounced from frustration over slow loading.

Here’s the equation for the money lost over the course of a year, where x = visitors and y = average revenue per customer, multiplied by 365 days, assuming the typical 2% conversion rate for ecommerce sites:

(.57x)(.02)(y)365 = $Lost Revenue$

The target?

Just three seconds. That’s the optimal window, after which a potentially staggering amount of revenue is at risk.

[You might also like: Why it takes so long for news, sports and travel sites to load]

The Good News: The Median Ecommerce Site Has Improved…

Our research for our last ecommerce report, Summer 2015 State of the Union for Ecommerce Page Speed & Web Performance, found the median shopping site approaching 2 MB in size, with about 170 resource requests – not so good, particularly when that resulted in a TTI of 5.5 seconds and a total load time of just over 15 seconds.

The findings in that last report were worse year-over year from 2014, and continued the trend towards larger, slower, and more complex pages.
Thankfully, there seems to be a break in the trend, which will hopefully be sustained moving forward.
Surprisingly, we found the median ecommerce site:

  • Has decreased in size from 1905 KB to 1395 KB, a 26% reduction in weight
  • Has reduced its number of requests from 169.4 to 97, an impressive 42.7% reduction, and
  • Has improved its TTI from 5.5 seconds to 3.1 seconds, a meaningful savings of 2.4 seconds and almost within the optimal three-second window.

In fact, 48% of the sites tested managed to make that window – an encouraging shift, and up from the 42% we saw the last time around.

Additionally, while the last ecommerce report found only a scant 3% of sites tested scored an “A” grade from WebPagetest.org for image compression, that number has increased to 10%.

These sites are employing web performance optimization (WPO) best practices, whether they are compressing and correctly formatting images or minimizing their use of JavaScript and limiting the number of requests and redirects.

Each one of these proactive measures reduces latency and streamlines the process of getting the webpage’s content to the user in an acceptable amount of time.

The companies with multiple regional sites that embraced web performance optimization best practices scored well almost across the board, with Amazon, Mercadolibre and eBay all posting impressive scores (although two of these companies had one regional variant outside the three-second window). A sizable two-thirds of the top-scoring sites were from these three companies alone.

Another standout was IKEA, which continued to improve from previous years.

In our previous ecommerce report, we found ikea.com had improved its TTI from 10.5 seconds to 3.6 seconds within six months’ time. Now, it’s almost at the very top, with an incredibly-improved TTI of just 1.8 seconds.

Looking at the waterfall, the efficiency of the page design stands out, with few requests to manage and everything loading quickly:

Even Amazon’s US site, which is magnitudes larger (4.17 MB to IKEA’s 258 KB) and more complex than IKEA’s (86 requests to IKEA’s 22) managed a TTI of 1.8 seconds – an impressive tie with IKEA.

If more companies can optimize their sites like these two, there should be a correlated uptick in revenue.

…But There’s Still Room For Improvement

While there is certainly cause for optimism for the positive gains in web performance, the battle hasn’t been won yet for ecommerce sites as a category.

Why?

  • Fifty-two percent of the sites tested still scored outside the three-second window
  • Thirty-six percent failed at image compression
  • Thirty-two percent of sites tested filed to cache static content, and
  • Forty percent of sites tested had more than 150 requests

At the bottom of the list, one site had 787 requests, was more than 9.4 MB in size and plodded along with a TTI of 7.7 seconds and a total load time of more than 30 seconds. It scored an “F” for both image compression and at caching static content, meaning elements had to be requested repeatedly.

Consider these charts:


Source: WebPagetest.org

The chart representing “Visual Progress” dovetails with its TTI of 7.7 seconds, and looking at the fact that the vast majority of its 787 requests are images, the culprit emerges.

Looking at the waterfall, you’ll notice that after the initial CSS and JavaScript, there are a long string of image requests. This continues unbroken for hundreds of requests in a row.

Recalling that each request takes up a server connection, and that there are a limited number of simultaneous connections available, all of these small images add up to one massive total.

It’s a bit like the expression, “death by a thousand cuts”: the little bits of damage add up.


Source: WebPagetest.org

This site is in dire need of a tune-up, and starting with image compression and “CSS Spriting” would be a great move.

CSS Spriting may be an older trick, but it could be useful in this case.

While image compression reduces the weight of the images, CSS Spriting groups multiple images together as one. As the Google Developers page states, CSS Spriting “attempts to combine all such images referenced from a CSS file into a single large image. The individual CSS backgrounds are then rewritten to point to the single large image, with background positioning declarations so that the page appears as it was before.”

In the case of the slow page described above, the multitudes of small images are killing the page’s TTI and load time. CSS Spriting could have a big impact by:

  • Requiring fewer sever connections in the browser,
  • Shaving off time by saving bytes from the overhead of each individual request (HTTP and Image Headers), and
  • Decoding one large image faster than multiple small ones can be decoded.

By addressing these issues, the site developers should see a noticeable improvement.

[You might also like: Why are eighty percent of travel sites failing customer expectations?]

Web Performance Optimization Shouldn’t Be A Tough Sell

In comparison to many other industries that more recently transitioned to robust websites, ecommerce has had a big lead, and it shows, at least anecdotally, in the results of our testing.

The relationship between the efficiency of the mechanism enabling the virtual shopping experience and the revenue generated by that shopping experience is relatively easy to deduce, compared to industries which are less transaction-based.

Avoiding site abandonment while simultaneously being able to process more transactions in a set span of time should be an obvious and powerful incentive, and yet we haven’t got the point where the majority of sites are nailing their TTI scores.

However, whether it’s due to the “carrot” of making more money or the “stick” of damaging the brand and losing customers, something appears to be going on with the overall web performance of ecommerce sites, and it’s encouraging.

Find Out More

For steps you can take to improve your site’s web performance and to get more analysis of the ecommerce sites we tested (in addition to news, sports and travel sites), read our article, “Performance Bottlenecks Evident in the Top 50 Sites of Multiple Industries” and get the full report, 2016 State of the Union: Multi-Industry Web Performance (Desktop Edition).

 

Remember – every second counts.

Matt Young

As a technology evangelist and writer for Radware, Matt Young delivers research and articles to the application delivery and web performance community. Before joining Radware, Matt was a top blogger for BlackBerry and he also served as the Web Editor for Avaya and as a freelance technology writer in the Greater Bay Area. Matt has a Journalism degree from San Jose State University.

Leave a Reply

Your email address will not be published. Required fields are marked *