Testing Retail Websites to Handle Holiday Traffic

It takes teamwork and the right kind of testing to prepare e-retail sites for peak holiday traffic.


Any online retailer who’s had a web site crash or slow to a crawl during a peak sales period will attest that time is truly money—downtime leads to lost site visitors, clicks and ultimately conversions. “What robs a web site of performance, steals from the bottom line,” says Joe Loveless, web performance market manager at web performance and optimization technology provider Neustar. “We see a direct correlation between a customer’s web site performance and revenue.”

Especially as the back-to-school season comes into full swing and the holidays approach, traffic-proofing an e-commerce site is a top priority if e-retailers are going to avoid losing sales. Doing that well requires a dual approach, Loveless says: Retailers should combine information from synthetic monitoring of site traffic with data about real customer site visits from various locations.

Synthetic monitoring provides a baseline measure of a web site’s performance regarding page loads and functions, he says, and retailers should monitor it regularly. In those tests, an automated cloud-based program sends out server requests to check how the web site loads from all over the world and reports whether all the page elements are operating as designed.

That’s an essential first step and it provides a necessary health baseline, however, it does not show a retailer how actual consumers are experiencing site loads. For a consumer from Northern Virginia using a fast fiber- optic connection versus a consumer connecting in the Philippines through a slower mobile connection, retailers need to understand the entire user experience. A synthetic-only approach also fails to account for other variations based on individual behavior.

For example, the site may load more slowly for a shopper whose past web activity or browser type prompts a lot of third-party content to load on the site, he says. “The needs of monitoring show us that an effective testing perspective requires more than just pounding away on a web site with X number of virtual users. Real browser loads are needed as well.”

Going Beyond Virtual

Those real browser loads require virtual user tests launched from a real browser, not just virtual machine simulations. Testing that way allows a retailer to pinpoint and fix irregular problems in advance of a busy traffic period.

When pre-production testing is completed, retailers need to keep a tight eye on how those operations are functioning. In one example, Loveless cites a customer who manages web services for large retail providers. A Pacific Rim-based e-commerce server could not be seen from the U.S., but testing and monitoring were inconclusive. Real user measurements were used to interrogate between the synthetic gaps. That testing revealed that a third-party marketing source, not included in the testing, was loading large videos caches on the web site, bringing about intermittent performance drags.

Performance Is a Joint Effort

As retailers prepare their sites for holiday traffic, they should keep in mind that the job of ensuring good performance is not just an I.T. department responsibility, Loveless says. It must be a joint effort between the technology, marketing and business teams, which have to keep each other informed of their plans, he says. Otherwise, the marketing team may create a promotion designed to drive 10% more traffic to the site on Thanksgiving, for example, but if the I.T. team doesn’t know about it, they won’t be able to allot the resources to handle the extra bump of visitors.

“To remain solely in I.T. runs the risk of missing some key objectives,” Loveless says. “The tighter retailers can tie the load test into marketing and business objectives, the better the web site is going to be able to perform to meet that.” More than anything, getting all internal teams to participate in planning for web site performance requirements is key for avoiding problems, he says.

Then, after retailers have made sure they are testing their sites from all angles and have accounted for all their business needs, they should always monitor how their sites perform during the peak period, Loveless says. Debriefing what did and didn’t work in one busy period will help online retailers to prioritize what to do to ensure optimal performance during the next peak period.