Engineers at AutoAnything.com, No. 204 in Internet Retailer’s Top 500 Guide, didn’t think the web site was particularly slow, but wondered if they could make it faster. The auto parts e-retailer maintains a deep inventory of 4.2 million SKUs and displays its home page with several dozen navigation categories and promotional prompts that could take 9.5 seconds to load for some visitors.
“The site was getting slightly bloated. As you get bigger pages, you lose focus and control on what the user experience is like,” says Parag Patel, chief technology officer at AutoAnything.
When the web acceleration firm Strangeloop approached AutoAnything and suggested it could improve the customer experience by speeding page loading times, Patel was open to listening. AutoAnything tested the service in January and February, but was adamant up front that any impact be measureable in terms of financial performance.
Being a car parts retailer, Patel described his demands for the service in car terms, says Joshua Bixby, president of Strangeloop. He says Patel saw the Strangeloop service as being like a side impact air bag in a car. “He’s said it was nice to have but that he could ultimately drive a car without one. We had to be like the transmission or the engine. We had to be something you have to have to drive the car.”
The service proved its value by splitting AutoAnything’s web traffic. It diverted half of site traffic to route through Strangeloop’s accelerator device before hitting AutoAnything’s server system; the other half of traffic went through AutoAnything’s servers as before. Strangeloop measured results using Google Analytics, the same platform AutoAnything uses to track performance. “We wanted to show a meaningful difference in the same terms they measured their own performance in,” Bixby says.
After two months of testing, the AutoAnything traffic routed through the accelerator showed an approximate 50% improvement in page load time, Patel says. The visitor-to-sales conversion rate for the accelerator-routed traffic increased by 9% while average ticket size increased 11%, compared to traffic not routed through the accelerator, he says. AutoAnything would not disclose its actual conversion rate or average order value. Traffic routed through the accelerator resulted in an increase in sales of 12% to 13%, Patel says.
Having these bottom-line results meant Patel could show return on investment and lobby for the budget to buy the service. “The appliance has obviously improved conversion and revenue for us. When you affect conversion by 9%, that is very significant on an annual basis,” he says.
After buying the traffic accelerator service, AutoAnything initially maintained a 95% to 5% split in traffic so it could continuously validate Strangeloop’s value to the site. But Patel says the firm recently routed the last 5% of traffic to move through the accelerator because it was too hard to ignore the temptation of the revenue it was missing by not using the accelerator. Both firms declined to disclose the service’s cost, but Patel says it represents less than 3% of the firm’s overall technology budget.