Mobile accounted for 25% of Ulta's e-commerce revenue during Q2.
Repeated benchmarking tests prove that less is more for TheMedicalSupplyDepot.com.
Meir Tsinman, president of TheMedicalSupplyDepot.com, says he realized a year ago that there was a lot he didn’t understand about his e-retail site’s performance. He knew conversions were low and bounce rates were high, but he didn’t understand why until he put the site through an intense, six-month review with the assistance of software vendor NetSuite Inc., which helped him understand and clean up the site’s performance issues.
The main issue, Tsinman told the audience at the Internet Retailer Web Design & Usability Conference Tuesday, was the site operated slowly, and one of the main drags on load time was the number of servers the site had to connect with to fully render a page. Before the review TheMedicalSupplyDepot.com had to make and process 80 requests from 16 outside vendors to fully render the home page. Outside vendors provided functions like user reviews and product comparison tools, which Tsinman admits are important tools, but not important enough to justify slow load times. He cited research showing conversion rates drop 7% for every second it takes a web page to load.
Tsinman wanted pages to load in two seconds or less, but that wasn’t possible with all the vendor tools he was using. “We had third-party tools that took a second each to load,” Tsinman said. “We had to ask: ‘Are those really important enough to have my customers wait to have the page load?’”
The site took a dramatic step a lot of e-retailers would be unwilling to make, said Andy Lloyd, general manager of e-commerce products at NetSuite. It removed every script from every vendor to see exactly how fast the site could load if it only had to download data from the retailer’s own web server. Then it slowly added one script back to the site and tested its page load speed. Then it removed it and tested the next one until it had a good grasp on what effect each vendor script had on page-loading time. “People have do a lot to their sites on a day-to-day basis, but are not consistently cataloging what is actually going on on their sites,” Lloyd said. “What we did here was track what was going on on the site and see what was affecting it.”
The retailer also realized it needed to make other improvements that would decrease load times. Tsinman’s team rewrote the site’s basic site code to make it simpler and sized down large images, for example. They also made sure to cache certain site features with web browsers so the browsers would not have to process new requests each time the features needed to appear.
At the end of the process, TheMedicalSupplyDepot.com cut the number of outside vendor scripts it embeds on its site to seven from the original 16. The site now makes 52 outbound requests, down from 80. It also worked with the vendors to reduce the amount of time it takes to load their code onto the retailer’s site. For example, a web site analytics script from Celebros took 984 milliseconds to load prior to the review; after working with the vendor, the time dropped to 250 milliseconds.
Tsinman did not disclose how long it takes for his site to load today, but did say key metrics greatly improved. Average number of pages viewed went up 20%, time on site went up 50%, the bounce rate improved by 40% and the conversion rate improved by 50%, he said.
Tsinman was careful to say that third-party scripts aren’t bad, just so long as e-retailers understand the trade-off they may be making for having them. “A vendor may tell you, ‘I can increase your conversions by 5%,’,” he says. “But if it takes a second or two to load, what help is that? Always choose the speed of your site, the performance, over the features. If a customer cannot land on that page, you can have all the beautiful features you want but no one will use it.” He says he will reintroduce some of the functionalities the site removed once it can do so without slowing site load times.