The online-only retailer grew sales by 11.8% in 2014.
New technologies help keep sites humming, and revenue flowing.
Speed matters. Online retailers know that sluggish site performance directly impacts the metrics they care most about: customer satisfaction, conversions, brand reputation and revenue.
When a shopper has to wait, she can quickly grow frustrated, making her more likely to click or tap to a competitor's site. And if that shopper finds the competitor's site meets or exceeds her expectations, chances are she'll remain loyal to that retailer.
There's a direct connection between site performance and sales. For instance, a recent Walmart.com study found that an improvement in page load time of 100 milliseconds—one-tenth of a second—produced a 1% increase in incremental revenue for the e-retailer.
"Site performance translates to financial performance for retailers, which is why they need to align their performance goals with their financial goals," says Joseph Loveless, market management, enterprise services, for Neustar Inc., a provider of real-time information and analysis, as well as web site performance monitoring and testing. "Poor-performing web sites frustrate consumers, who then visit and purchase from those sites less often."
Because site performance can influence both a shopper's perception of a retailer's brand and, even more importantly, whether he makes a purchase, retailers have to measure site performance in real time from a shopper's point of view. Unlike synthetic performance monitoring, which simulates visitor traffic to determine web site performance, real user measurements collect actual performance data from a visitor's browser, which is sent to a back-end server for analysis. To gather the data, the retailer attaches a web beacon to pages that it wants to closely monitor, such as product showrooms and shopping carts. Beacons are small tracking GIF files that gather more information than cookies and can help detail more precisely a consumer's performance experience.
"Real user measurements provide a deeper look into site performance, allowing retailers to catch problems that can be missed with emulated monitoring," says Loveless. "The two forms of monitoring complement each other and provide retailers with a more complete view of site performance."
A balancing act
While speed is a critical element of site performance—even mobile users expect web pages to download in less than 2 seconds according to Google Inc. research—that can't be web retailers' sole focus. Consumers not only demand fast interactions from the retail web sites they visit, they expect an engaging shopping experience. But showroom features like 360-degree product views, zoom and large images slow page access times. That's why retailers have to balance the experiences they offer with the speed consumers expect.
"Retailers have sacrificed a rich interactive shopping experience in their quest for faster performance, especially for mobile sites," says Peter Blum, vice president of product management for Instart Logic Inc., which provides a cloud-based acceleration service called the Web Application Streaming Network. "E-retailers using our web application streaming have been able to deliver the speed and the engaging shopping experience that customers respond to. And their conversion rates have risen."
For example, Dollar Shave Club, an e-retailer of razor blades, saw conversion rates increase more than 16% after installing Instart Logic's acceleration service, says Blum.
To achieve that result, Instart Logic replaced downloading with streaming technology and added intelligence about the content being delivered. A content delivery network replacement, the Web Application Streaming Network contains knowledge of how web site and application components are consumed at the browser level. It splits the content or page into components and first streams—rather than downloads—the highest priority pieces needed to display a web page so a shopper can start interacting with it. Once the initial content is received at the browser level, the remaining portion is streamed in the background.
This method is significantly faster, on both mobile and wired connections, than downloading every element on a page before a user can interact with it, Blum says. It is similar to what occurs when consumers stream video and music: The beginning of the video or song is streamed so it can immediately be played and enjoyed, while the rest of the content streams in the background. In contrast, content delivery networks download all page elements on an e-commerce site before a shopper can fully experience the page.
"Our approach is to replace existing content delivery systems with intelligent web app streaming to allow retailers to create visually rich—even immersive experiences—without having to worry about the performance for the end user," Blum says.
Because so much information must travel back and forth between a retailer's server and a consumer's browser to render a page, retailers must be able to quickly identify bottlenecks when site performance slows. Two common culprits for performance issues are defective page coding and excessive web server traffic.
Using real user measurements makes it easier for retailers to quickly determine whether performance problems are due to page coding issues as opposed to sudden spikes in traffic, because they take exact measurements of timed performance with real users. In contrast, emulated monitoring is less exact, because it relies on one machine asking another how it's performing in timed intervals.
"The ability to measure all the components of a page in real time enables retailers to look beyond problems caused by traffic overloads and spot where the actual performance bottlenecks are so they can address those issues faster," says Neustar's Loveless. Load testing—which determines how long it takes to complete the dozens of round-trips between the server and the browser to retrieve such page elements as style sheets, scripts and images—takes place in a simulated environment in which a web server is bombarded with a sudden increase in page requests to determine the upper limits of traffic it can handle. This information tells a retailer how much additional server capacity it needs to handle peak traffic.