The old rules for speeding up ecommerce websites are toast. That’s the clear conclusion to draw from a recent Pingdom study comparing performance of the top 100 ecommerce sites—including dozens of big retailers’ sites.

Some fat sites are fast anyway. Some lightweight sites are surprisingly slow. And what’s really killing performance seems to be metrics. The good news is that virtually all the large retailers got their response times down under the fabled three second mark. The bad news is there’s no longer a clear correlation between speed and site size and the number of files requested, the variables that Pingdom tracked in this study.

The chains with the fastest sites—we knew you were waiting for this—are Forever 21 (#2 out of 100), Ikea (#4), Walgreens (#5), REI (#6), Kohl’s (#7) and Bloomingdale’s (#10). Bringing up the rear is Neiman Marcus (#92), which along with Victoria’s Secret (#89) and Bed Bath and Beyond (#86) had the only sites that took more than three seconds to load.

Among other major chains, Walmart (#57), Target (#60) and CVS (#81) were top-10 chains that fell in the bottom half of the rankings, behind Best Buy (#13), Lowes (#24), Home Depot (#28), Sears (#32) and Kmart (#39). Amazon was #20. (A link to the complete list is here.)

But enough of the horse race. What’s much more interesting about the results is how random they appear, at least according to conventional website performance wisdom. Pingdom headlined its survey report ‘Slim and trim to slow and bloated—the top 100 E-Commerce websites,’ but that’s a little misleading. Two of the fastest sites (REI.com and Kohls.com) are also among the most bloated, as measured by the amount of data that has to be downloaded—they’re more than twice the size of an average site in the study.

That’s not the only place the old smaller-is-faster paradigm falls apart. Yes, Ikea has a slim site that’s also fast. Forever 21 is even faster, but it comes in at #62 for slimness—it’s heftier than average.

Meanwhile, Saks Fifth Avenue has one of the smallest sites, but it’s only #49 for speed. And CVS, at just over half the average size, is simply slow compared to almost all the other retailers.

True, some of the smaller sites perform faster. But smaller-is-faster is no longer a reliable rule – if it ever was.

Ironically, what’s slowing down some of the big-chain sites are services that rely on third parties—such as web metrics and tracking services—that drag down site load times but are outside the control of retailers. Ironically, what’s slowing down some of the big-chain sites may not be problems that retail IT shops can easily fix. Inadequate bandwidth and not-fast-enough servers might be part of the problem. But a clear reason some sites can’t finish loading quickly is that they are calling out to third parties.

That’s not in Pingdom’s study, but a little digging with Pingdom’s own tools makes it apparent. For example, the slowest big retailer site NeimanMarcus.com is actually 10% smaller than top-ranked Forever21.com. But almost all the HTTP requests for Forever 21 actually go to the retailer’s site, while a quarter of Neiman Marcus’s HTTP requests go to outside sites—including one that, in a test run we made using Pingdom’s tool, took more than a second and a half to deliver a few hundred bytes.

That’s what kills Neiman Marcus’ response time. Or rather, it’s what slows the site’s speed down so it loads in under four seconds instead of under three seconds. That’s still fast—probably fast enough that customers don’t care. The must-load-in-three-seconds-or-customers-will-leave rule is also highly overrated—it’s based on studies of mainframe users from decades ago—but three seconds remains the target for most ecommerce sites.

And that three-second target is probably the biggest challenge any ecommerce team faces in trying to improve site performance. It’s pretty clear that if a relatively fat site can load in less than a second, the idea that sites have to be slim to be quick is dead. The fattest site in Pingdom’s survey, which was Ticketmaster’s site at 2.5MB, loaded in 2.33 seconds - well under that three-second target. Clearly, all that effort to speed up sites really has paid off.

Of course, once a site loads in less than three seconds, performance is obviously fine—which serves as an invitation to designers to add more complicated graphics, to CRM specialists to add more tracking tools and to anyone else connected with the site to fill up the bandwidth. If there’s any time to spare, they’ll fill it.

That’s not a bad thing in itself - CRM data is useful, and if fancy graphics make products look more appealing on a high-definition display, so much the better. But customer tracking and other functions will always end up soaking up more performance than is actually available. That means the site has to keep getting faster, and any headroom will attract even more speed-devouring ideas.

From here on in, keeping your site fast is going to require a lot more fine tuning—and you’ll get a lot less help from rules of thumb. The simple ways of speeding up sites have pretty much been used up.

StorefrontBacktalk

StorefrontBacktalk is the U.S.’s leading source for global retail technology, e-commerce and mobile commerce issues, including security, in-store strategies and CRM issues.

Visit StorefrontBacktalk.com or StorefrontBacktalk’s Kindle version or Twitter feed for the latest insights and information.