For better or worse, the Web is perhaps the most infinitely measurable marketing medium we’ve got today. Marketers have access to tools that can tell them who’s coming to their Web sites, where they’re coming from and how long they’re staying, what they’re looking for, how much they’re spending and how often. That kind of visibility is one reason a recent Fortune magazine cover story suggested that big brands are mulling a move from TV advertising to the Web: the ability to tell who they’re reaching and with what effect.
But strangely, that measurability can also be a hindrance to selling well on the Internet. When you can put a tape measure or stop watch to so many variables, how do you know what metrics are important for your specific situation? As a result, many e-commerce merchants shy away from anything more than the most rudimentary Web analysis; other, braver ones may dive in and quickly get swamped in the streams of data that can come from making a few simple changes to a Web site.
Josh Manion, CEO of Web analytics firm Stratigent, says he’s seen both these approaches to Web analysis scuttle merchants’ attempts to optimize their Internet performance. In some instances, he says, site operators mistrust Web-based data, perhaps in a holdover from the old dot-com days when statistics could be used to prove that you could make millions selling dog food over the Internet.
Then there are others who want full access to the raw Web results generated by firms like Manion’s. “We’ve had very limited success with giving clients full access to this data,” he says. “They start wandering around among 300 possible reports, they don’t know what any of it really means, and they just start asking questions about issues that aren’t really relevant to improving their Web site performance,” he says.
Instead, Manion advocates a rigorously scientific approach to conducting a Web analysis. He’s far from alone in that, but he is perhaps uniquely emphatic in insisting that the whole analytics process must proceed from careful definition of problems, through hypotheses about what’s causing those problems, to controlled testing of variables, studying the data from those tests, and drawing conclusions about whether those changes solved the problems they were meant to address.
“Too many operators make changes to their Web sites based on irrelevant outside factors, such as that the Web designer likes it, or even the CEO likes it, or someone did it this way at another Web site,” Manion says. “People wind up doing things that aren’t valid based on either intuition or past learnings that may not apply to their case. Data trumps intuition,” he says, quoting Ronny Kohavi, director of data and personalization for Amazon.com.
In Stratigent’s view, scientific Web optimization starts with a baseline examination of current Web site performance to set some benchmarks for measuring enhancements. At this stage, key performance indicators are identified, the metrics that will be watched most closely—whether conversion rates (sales or other actions), aggregate measures like Web traffic, or components such as registration or checkout processes.
Next, firms such as Stratigent design focused “experiments” that change one variable or a managed set of variables on a Web site. Manion says this is the stage where some operators load in other ancillary Web projects such as repairing a poor shopping cart. Those changes alter the experiment, he says; stay focused on the problems identified at the beginning, or your results and conclusions may not point up the right long-term solution for your Web issues.
And be sure to compile enough data to make those results statistically valid. That means watching how a single Web change performs with a lot of traffic, and over a useful period of time. “If you have a high-traffic site and run a test for one day, you may get a lot of volume for your results,” Manion says. “But a Monday, for example, may not be representative of your average Web business. You need to let the test run long enough to normalize the data.”
Although this scientific process may sound intimidatingly clinical and lengthy, it doesn’t preclude finding some quick fixes in the initial site assessment. Designer Linens Outlet, a Stratigent client that sells excess inventory from a large linens wholesaler, turned to Stratigent in 2004 for help with three primary issues: improving the effectiveness of online marketing, optimizing product placement on the site, and boosting conversion rates—including both sales and leads from registration for a newsletter.
Manion says the initial assessment of the Designer Linens site turned up something that helped with that last problem: The newsletter registration process was incompatible with some versions of popular Web browsers. The issue was quickly fixed, and newsletter registrations grew by about 270%. “Sometimes we come across that kind of low-hanging fruit,” he says.
With those easy moves done, a process of ongoing site-wide optimization began, and Stratigent began designing and implementing experiments to test and measure possible solutions for Designer Linens’ problem areas. For example, the Web site was using a double opt-in newsletter registration. Under Stratigent’s guidance, it tested a single opt-in process and found that registrations grew while unsubscribes remained flat. That change brought registration conversions up even further, to a total 400% increase.
Another experiment tested the effect of adding broader shipping options to the Designer Linens checkout process. That change, it turned out, fostered a 20% increase in conversions from the shipping page to the payment page.
Other tests examined the conversion rates for specific Designer Outlet products. “That’s important to me from a merchandising and marketing standpoint,” says Beverly Dantz, interactive marketing manager for Designer Linens outlet. “If I want to promote certain products in our campaigns, I want to know which ones are the most likely to convert.” For example, Web analysis revealed that while one product line, the Memory Foam mattress pads and tops, was not among Designer’s five top-selling items, it was one of the five leaders in conversions. That line was brought forward and made easier to find on the home page.
Designer Linens recently outsourced its search marketing effort, but Dantz says that before that step, she also used both Stratigent’s product conversion reports and Web analysis applications from WebTrends to pinpoint the optimum keyword bids for the highest return on that investment. “We ran unique tags on each keyword in a campaign to measure ROI,” she says. “That allowed us to figure out what we should be bidding for those terms. You can spend up to $5 on some of these words, but in reality sometimes they’re not worth 20 cents.”
Designer Linens Outlet now has Stratigent on retainer for continuing analysis. One reason is that the company’s Web site is in the middle of another round of changes. But beyond that, Dantz says her small marketing team finds value in the distilled analysis reports that they get from Stratigent. “I know the power of the information that’s in the WebTrends platform, but I don’t have time to pull it myself,” she says. “It’s easy to fall into information overload. Stratigent helps us focus on the key performance indicators that we really want to track. They generate a monthly report for us showing how we compare to the same metrics the previous month, and that keeps us on track.”
Manion says that Dantz’s situation is a typical one. “They’re empowered by using our reports in combination with the WebTrends analysis tools,” he says. “They can check stats daily during campaigns, and we produce a detailed report that we go over with them once a month.” His firm can produce a portfolio of dashboards that give different elements in a company—the marketing team, C-level executives, content developers and such—the current stats in the format that’s most useful to them.
And e-tailers of all sizes are coming to view Web analytics as an ongoing process that’s crucial to their success. “Two years ago, many people thought Web analytics were just for the big names on the Internet,” Manion says. “But now, not taking advantage of Web analytics is coming to be seen as putting yourself at a competitive disadvantage.”
Still, Stratigent encounters companies at all levels of e-commerce that are reluctant to commit wholeheartedly to a continuing careful analysis of their Web site. That may partly be resistance to adding another operating cost. But the complacency factor may be just as important. Some of these companies have migrated to the Web from other channels, and while they might not dream of sending out a catalog or direct mail piece without testing, they are content if their Web store seems to be functioning at less than peak efficiency.
“I’ve spoken with merchants on the Internet who say that their site brings in 20% of their annual revenue without any tweaking, so why mess with it?” he says. “They say they’re getting a 37% conversion rate, and that’s good. I find that very frustrating: It’s like saying, ‘I’ve got $100, so I don’t need any more money.’”