Web Analytics — What's Really Important?

The sheer volume of data chronicling the activities of your online shoppers is enough to make your head spin and your servers overflow. Who hit what page, where they were shopping from, what they were shopping for, when they were shopping for it, where they came from to look at it, and where they went when they quit looking at it are easy questions to answer about online consumers. But which of these are most important to know, what else can you learn about your Web shoppers, and, most importantly, what can you do with the intelligence you gain on your Internet shoppers' behaviors?

Can Anything Beat T-Log Data?
Mark Goldstein is what I would call a pragmatic proponent of Web analytics. The Loyalty Lab Inc. CEO recognizes the value of garnering intelligence on your customers (as any CRM [customer relationship management] software company CEO should), but he refuses to discount good old-fashioned T-log data as the best information you can have. "Mining a customer's transaction history and learning his purchasing habits is more powerful than any other analysis you can do," he says. Fortunately for retailers — online or otherwise — T-log data is also the easiest to access and the simplest to use. Quite simply, if you know what customers buy, you know what to promote to them. But Goldstein is careful to point out that in T-log analysis, longevity of analysis — in terms of years — is key. "You can act quickly on data that tells you what your customer bought in the last few days or weeks, and that's important," he says. "But analyzing purchase history year over year can be lucrative too, because that gives you a look at seasonal and holiday spending patterns," he advises. There's inherent value in knowing that a customer buys pool equipment and accessories every May and ski apparel each November, and you won't get this big of a picture if you only analyze short-term T-log data.

What's Better To Know — What's Selling Or What's Not?
While Goldstein's contention that T-log data is valuable can be lost on no one, there are those who would argue that there's at least equal value in knowing what shoppers didn't buy. This is where Web analytics solutions really shine. In short, these are solutions that allow analysis of customer shopping — as opposed to buying — behavior. Web analysis allows retailers to see click-thru rates (frequency of visits to specific URLs [uniform resource locators]), basket additions and subtractions, time spent browsing specific pages or products, and what pages consumers are visiting before and after yours, to name a few. Joe Davis is CEO of Coremetrics, a major player in the Web analytics space. He says that collecting data on merchandise that shoppers are putting in their shopping baskets but ultimately not buying is valuable. "Monitoring this kind of activity can give you insight into trends and patterns that might indicate overpriced merchandise or better deals elsewhere," he says. But having this insight goes from valuable to actionable when that abandoned merchandise can be attributed to a specific individual. "If you can capture an e-mail or mailing address, you can create much more targeted marketing efforts. For example, you could send a targeted promotion of your most-abandoned merchandise to those specific individuals who were so close to buying it," says Davis. For multichannel retailers, this could be a particularly lucrative strategy. "The Web is really important to retailers because it's a means of preshopping," says Goldstein. "It's estimated that 60% of Sears customers preshop online, then go to the store to buy." Web analytics make it possible to identify those shoppers and know their intentions before they even set foot in the store. What could that kind of information do for your targeted marketing campaigns?

Of course, cross-channel marketing execution is dependent on CRM systems that provide a single view of the customer regardless of channel. Historically, eliminating silos and gaining a comprehensive view of customers across channels was extremely challenging. "Cross-channel integration used to take millions of dollars and prayer, but through Web services you can more easily integrate disparate data silos," says Goldstein. "Traditionally, the only option was to license an enterprise CRM system and re-architect the silos to build one infrastructure, but XML [extensible markup language] allows us to integrate the silos together now."

E-Tail Lessons Learned
The instant-feedback nature of the Web is great testing ground for multichannel retailers to determine merchandising strategies that might work best in stores. "Online, you can analyze the effect storefront changes have on sales within hours," says Davis. "Multichannel retailers can therefore use Web results to be more effective with merchandising strategies in other channels."

Web testing can also help retailers make merchandising, sale, and clearance decisions. For example, apparel retailers can inexpensively gauge the sales of certain styles and colors online before making expensive store stocking and catalog printing decisions. They can also test price changes online and evaluate the effect of price changes on merchandise movement before implementing store-level clearance sales. Before you had a Web site, you had to make these pricing and merchandising decisions, implement the changes, then play the waiting game before you had visibility into what was selling and how well. Regardless of which data you find most valuable, Goldstein says rigor is the key to making things happen with Web analysis. While Davis will be happy to tell you about new, sophisticated solutions that do the analysis and make merchandising suggestions or even automate response actions (including marketing campaigns) based on buyer behavior, Goldstein says the human element is still key. "Retailers may be proud of their contracts with [Web analytics companies] Omniture, Webside Story, and Coremetrics," he says, "but they're not necessarily using them to their fullest potential. I would advise retailers to get someone in-house that has a full-time data management and evaluation job."