Web analytics are nothing new. In their most familiar form, you get statistics reflecting collective behaviors—numbers of views, clicks, conversions, etc.—that help you optimize ads, media buys, search terms, and overall site design.
That’s all well and good—but you can do better. “Predictive” analytics on Web data drill deeper, tapping individual customer, prospect, and visitor insights so that you can target and personalize your Website and your ad and e-mail campaigns for much greater impact. By linking individualized data on search terms, geolocation (based on IP addresses), registered customer information, and the like to actual behaviors (views, clicks, orders), you can address a key marketing question: Which type of person is most likely to be interested in which of your services, products, or content?
The “type” forms a profile. The more closely your Web visitor conforms to a profile your predictive analytics have discovered (as reflected in scores), the more likely that visitor will behave in a manner you can anticipate.
But although many Web marketers have heard about predictive analytics, too few can explain their practical value. Here are four field-proven ways you can apply predictive analytics to empower your online marketing in ways you may have only dreamed about before.
1) Recommend products. If you’re like millions of other consumers worldwide, you’ve experienced the product-recommendation function on sites such as Amazon and Netflix. In concept, the principle is simple: The site presents suggestions based on what people who looked at similar pages (or bought similar products) have also viewed or purchased.
Execution, however, is more complex. The greater the number of products involved, the more elaborate the predictive calculation. Recently Netflix offered a $1 million prize to anyone who could demonstrably improve its recommendation function. A collaborative effort among three development groups won the money—but Netflix was a winner as well, because the new prediction capabilities resulted in a 10% improvement in prediction quality over its previous model.
2) Allocate retention dollars. You know the marketing truism: it’s much cheaper to retain a customer than to acquire a new one. But on whom should retention dollars be spent? If a customer is likely to stay without prompting, any retention budget spent on him is money wasted. Worse, there’s an entire category of customers who might respond with the behavior you don’t want: leaving.
Within your database, customers fall into four distinct groups:
• Lost Causes, who will leave regardless of your efforts
• Sure Things, who will stay regardless
• Sleeping Dogs, who will leave as a consequence of your promotion
• Persuadables, who are most likely to stay as a result of your retention efforts.
Using predictive analytics, you can direct your retention dollars where they belong: to the Persuadables you can keep by giving them an additional incentive. We’ve found that adding a layer of Web behavioral data to existing efforts generally provides a 10%-15% lift over traditional churn modeling.
3) Select the most-appropriate ads. Suppose online ads and other marketing messages could be as precisely targeted as product recommendations? By applying predictive analytics to visitor profiles, you can deliver more-relevant messages and increase the depth of engagement.
In a recent case, a leading search service for colleges and student loans put its ad program to the test. At stake? A bounty of $25 for each information-request form visitors submitted. By comparing visitor behaviors with previously compiled profiles, the site was able to make educated guesses about which sponsor’s interstitial ad different visitors would find engaging.
In an A/B test of the site’s legacy system based on aggregate data versus a new model leveraging predictive analytics, the results were staggering: The predictive model increased the information-request rate by 25%, resulting in a 3.6%-5% increase in revenue, or about $1 million in additional revenue every 14 months.
4) Send follow-up e-mails for cross-sales or remarketing. The logic sounds simple: Why not send targeted e-mail offers to visitors who look at, but do not purchase, products on your site? In real life, however, many visitors may regard these e-mails as spam, so you need to carefully target the messages to those most likely to be receptive.
Kim Larsen at Charles Schwab conducted a study comparing the impact of follow-up e-mails against a control group that received none. The result was a wash: During a 90-day period, both groups had the same purchase rates.The problem was not that the e-mails failed to have impact; the issue was that the messages in the study were untargeted and offended as many customers as they inspired, negating any positive influence the campaign might have had.
To improve effectiveness, follow-up e-mail campaigns should segregate prospects into groups similar to those identified in customer retention programs. Only those prospects with “persuadable” profiles—as matched through predictive analytics—should be contacted, resulting in higher yields.
Integrating analytics and automation
Neither predictive nor web analysis by itself generates profits. Marketing automation is the key to turning insights into action on an industrial scale—in other words, in a repeatable, cost-effective manner. Bring the results of your analytics to fruition by integrating all of your efforts into one suite of activities that can be easily managed and monitored for maximum effectiveness:
• Web analytics with a built-in visitor-level data mart feeds predictive analytics with the necessary granular information.
• Predictive analytics solutions automatically score current Website visitors based on models derived from past visitors.
• Interactive marketing solutions combine visitor, customer, and predictive score data with marketers’ business rules and self-learning intelligence.
• Execution solutions deliver the message to whichever inbound or outbound interaction point your customers use.
By completely integrating Web and predictive analysis from end to end, you can close the marketing loop to capture the true, full value of every customer interaction.
Akin Arikan is director of product strategy and web analytics evangelist at Unica and the author of Multichannel Marketing: Metrics and Methods for On and Offline Success.
