The sky is falling: leading e-commerce sites degrade customer experience

Last week, “IT Pro” emphasized that “APM pays off“: market-leading e-commerce sites are slow enough to annoy consumers, even though well-known techniques have the potential to improve site responsiveness markedly. That contrast calls out for prompt attention.

It deserves immediate attention because those sites are not just slow, but getting worse, according to further measurements provided by Joshua Bixby.

Bixby aptly expresses the findings in terms of three “myths” he hopes to bust: pages are getting faster, users are satisfied, and browsers can make up any consequential deficiencies. He’s absolutely on-target. The one analytic gap I see in his myth-busting is insufficient emphasis on how the main measurements of the report likely underestimate the real-world impact of slow, bulky page loads. Most of the growth in use of the Web now is with mobile users, and in new specialties such as gaming sites. End-users of the latter are even more demanding than average for quick page updates. At the same time, mobile end-points are so variable that I’m convinced no one yet understands their real-life page-load profiles with much precision. If desktop browsers load average Web sites too slowly, displays on cellular handsets and tablets are likely to creep.

With great trouble comes great opportunity for improvement. A surprising number of sites can significantly improve their page loads with a few simple technical fixes such as HTTP compression and JavaScript minimization. More formal application performance management (APM) tools not only help track down chronic “hot spots” that drag down performance, but also save DevOps staff time that otherwise would go to “manual” diagnosis.

The sky is falling for the many Web sites which frustrate and annoy would-be users with their pokey page-loads. Most of them, though, can use APM and good Web technique to restore healthy, fast operations. Much of what is missing is simple recognition of the importance of load times.