eCommerce companies are acutely aware of how much page load times can affect their bottom line.  They know that milliseconds matter on the web.  Unfortunately, many eCommerce companies have been misled on the subject of how to improve web performance. At Yottaa we hear the same story again and again: an eCommerce site has migrated to a new, ostensibly faster platform at great cost to productivity and time for their web team – only to find that they are still plagued by the same performance problems.  These companies are the victims of the Platform Migration Lie.
 

eCommerce Today: The Performance Bottleneck Has Shifted

The Platform Migration Lie is rooted in the common misconception that the backend system powering an eCommerce shop is the primary driver of user experience on the site.  The idea seems logical, after all: the ecosystem of servers and databases handling requests and secure transactions is at the root of every eCommerce shop.  Without those components working properly, the shop could not operate.

But this datacenter-centric view of the web is outdated.  Websites today, especially eCommerce sites, are loaded with heavy images, JavaScript plugins, and social media integration.  This complex composition leads to 80-90% of the load time of websites being consumed on the front-end – the visitor’s browser – since it requires so much work to download and render the content.  That leaves the backend components like databases and servers to account for a fraction of performance.

The solution to faster pages, therefore, is to optimize the front end.  Sure, you need an eCommerce platform that can handle your requests and transactions efficiently at scale.  But if you’re going to seriously improve page load time, eCommerce sites need to optimize images, JavaScript and the other components on the front end.

“”We’re The Fastest eCommerce Platform Available!””

Here’s where the “”Lie”” component comes in. Salespeople at platforms try, and often succeed, to persuade eCommerce shops to migrate based on speed as a differentiator.  They prove their platform’s superiority using a simple comparison: “”This is your site now; this is your site on our platform.””

These comparisons are wholly problematic.  They neglect to control for the factors that are most important to performance today: the content of the page and the variability of a distributed audience.  What results is an apples-to-oranges comparison.

Here’s a common scenario.  Below is performance data from a typical eCommerce site – we’ll call it BobsHats.com. With a Time To Interact of nearly 8 seconds performance is slow, but it’s actually not far from the eCommerce industry average of 7 seconds.  Note the high count of images, JavaScript files, and other content types.

Seeing an opportunity to optimize his business by improving his page load time, Bob takes a sales call from an eCommerce platform that promises it’s the fastest on the market.  To prove that claim, the salesperson takes the data from a handful of Bob’s products and uses them to populate a page running on the new platform. The page looks roughly like an eCommerce page, but there’s a conspicuous absence of JavaScript widgets, banners, and far fewer images than the norm.

A performance test on this sample page turns in the following results:

The performance is outstanding.  It loads in just over 2 seconds. But it’s not because the platform is necessarily faster (it may or may not be) – it’s because there’s hardly any content on the page!  Since 80-90% of load time is consumed on the front-end, a site with hardly any content is always going to load fast, regardless of the platform.

To underscore the difference, here are the stats from the backend performance – the part that’s actually affected by the platform.

On the left is the original site, the right is the salesperson’s sample site.  Clearly, performance is better on the new platform.  But notice that this accounts for 0.5 seconds of load time, while the difference to overall load time shown in the user experience tests was a far more impressive 6 seconds.  

What the salesperson won’t admit is that 5.5 seconds of that 6-second difference is due to factors outside the control of the platform itself.  Moreover, this page is being sourced by a nearly empty database that’s optimally configured, so even the 0.5-second improvement seen here could be better than a real-world scenario.

But Bob doesn’t know all that.  Impressed by the raw comparison, Bob migrates to the new platform.  Once BobsHats.com is live on its new home, the bevy of hundreds of images, JavaScript plugins, and third-party integrations is back online.  The pages are once again heavy, and the database is full.  Average load time goes back up to the 7-8 second range, the way it was before. The page load sequence is distributed like this: 

Bob’s time, resources, and money have been wasted: he’s been successfully duped with the Platform Migration Lie.

Don’t Be A Bob

The main reason the Lie persists is that platform migration is seen as a solution to eCommerce problems writ large.  Realistically, changing your platform can only solve a narrow set of issues, the biggest of which is scale.  (If you outgrow your current provider and need a more scalable solution, then it’s reasonable to migrate).  But if scale isn’t an issue, and you aren’t experiencing obvious “”dealbreaker”” problems such as routine crashes, chances are you can achieve the improvement you seek using web performance solutions.

The key steps for eCommerce owners seeking better performance:

• Go to WebsiteTest.com and determine the key bottlenecks on your site with thorough performance testing
• If the key bottlenecks are content/front-end-related, seek a performance solution tailored to those needs (like automated optimization from Yottaa or manual optimization by your front end team)
• If the platform does seem to be a bottleneck, be meticulous when considering migration: request that platforms create a page that’s an exact copy of one yours, including all on-page assets, and independently test each with equal criteria.