A pet supplies retailer avoids accidents with pre-launch load testing of a new e-commerce site
In business since 1979 and online since the mid-1990s, pet supplies and medicine retailer KV Supply knew exactly the page elements it needed when it came time to redesign its more-than-a-decade-old website. It just didn’t know how the new version would hold up to real visitors and variable traffic volumes, says Dan Metzner, KV Supply’s web developer. “When it came time to think about launching it, we had tested the functionality but we didn’t know exactly what would happen when we unleashed the world on it,” he says.
As it turned out, there was a problem with the new version of KVSupply.com. The shopping cart would slow to a crawl if too many visitors added and removed items at once. But KV’s customers never had to deal with that frustration because the retailer caught the problem before launching the new site in tests with web performance and optimization technology provider Neustar Inc.
“We were afraid we were going to lose revenue—from years of experience we know when the site doesn’t perform well, people just leave,” Metzner says. “We didn’t want to lose any existing customers or scare away anyone new, especially because the new site looks and feels and works very different from the old one, so it was well worth waiting rather than launching something we weren’t certain of.”
KV Supply could hardly risk a falloff in web orders because of poor site performance. About 75% of the company’s orders and 60% of its revenue come from the website, Metzner says, with the rest coming from mail order or store sales.
In order to test how the new e-commerce site would perform when actual consumers were using it, Neustar wrote several programs that mimic the action of website visitors adding and removing items from their shopping carts and going through checkout. It simulated traffic in batches of simultaneous users, ramping up the traffic load on the site by 50 at a time up to 300 total concurrent users, Metzner says. “We noticed right away that as soon as we had about 50 people on there at once, the CPU usage was spiking up at 80%,” he says. In other words, the server hosting the site was overloaded, and page load times slowed markedly as a result.
Metzner, who built 90% of the new website himself, worked with his Neustar engineer to pinpoint the specific pieces of the checkout process causing the problems. He then spent two to three weeks fixing the bugs, and Neustar then retested to make sure the changes worked, he says. The revised shopping cart successfully processed 63% more transactions than the first version.
Neustar also tested how KV Supply’s online prescription ordering page would function with lots of site visitors. The test identified some dead links or page errors Metzner says he wouldn’t have found himself before the launch. Finally, Neustar verified that KV Supply could move the site off its in-house server to an external host that Metzner had selected without impairing perfor- mance. In fact, the test showed that the site’s pages loaded a bit faster when hosted externally, Metzner says. Each test took about an hour to run, and KV Supply needed to give Neustar only a day’s notice to run a new one, he says.
With all the corrections in place and Neustar’s assurance that everything was working properly, the retailer launched its new site in March. It hasn’t had any performance problems yet, Metzner says, even though web traffic is growing steadily. In May, visits to the site were up 40% from the same month last year, he says.
“It gave us peace of mind, a picture of what it would be like if we launched,” Metzner says. “Rather than using our customers as beta testers, we’d know it wouldn’t crash right away.”