Most business people these days are trying to read minds — they want to predict their customers’ individual needs and behaviors.
It’s rare to read a business publication without finding at least one article about customer relationship management — with everyone promising one-to-one personalization. Personalization — tailoring content to individuals — is becoming the holy grail of e-commerce.
Everyone with customers and online visitors wants to provide engaging, personal content to draw people back and support an objective – like selling more. How do you find your way through the avalanche of advice and implement a solution that builds more profitable customer relationships through personalization? The answer lies in understanding the different personalization methods available.
The most common means of providing personalized content is through a recommendation engine. A visitor makes a series of choices and, based on those choices, a recommendation as general as a Web banner or as specific as a product pitch is made.
What mechanisms are used to do what humans do so naturally: make a suggestion?
The most common form of delivering personalization is through business rules. Business rules are “if-then” statements based on common sense and conventional wisdom. For example, we know that women purchase women’s shoes and men purchase men’s shoes. Using business rules, an online merchant should recommend only women’s shoes to women. The downside of this method is that it does not look at the customer’s individual buying history or buying patterns within a population or group of shoppers like her.
What if a female customer has purchased shoes for herself and her family, but the online retailer only offers her women’s shoes? That retailer is not only missing the opportunity to increase the sale, but is withholding information she might consider valuable.
Another problem with relying on business rules for personalization is that as more customers visit the site, more rules need to be created and maintained over time, which is time-consuming. As more rules are added, there is a higher probability that those rules will conflict, leading to customers getting the wrong recommendation. In the previous example, the retailer may develop rules to recommend children’s shoes to women, but what if the customer has no children?
Since business rules are based on the experiences of the person who writes them rather than the history and behavior of the customer, they merely mimic personalization. Effective personalization, delivering valuable content to an individual, must consider the previous behavior of that individual. For example, a 40-year-old man and an 18-year-old man visit an online music store to buy the same Frank Sinatra compact disc. According to business rules-based personalization, they should both be interested in purchasing a Dean Martin CD. However, these customers are not alike, and they are probably purchasing these CDs for different reasons. They should receive different recommendations for their next purchases.
Data-driven personalization offers an alternative. It enables Web sites to tailor content to customers according to the actual behavior of individuals and populations. By analyzing individual customer behaviors and customer population behavior, an online merchant can understand what his customers want and predict what they will do.
Personalization helps understand and predict consumers by developing a complete picture of each customer, which means collecting and analyzing individual customer data in addition to demographic and transaction data. This information, combined with the company’s product information, creates models of behavior to segment or classify customers and buying behaviors. However, to get a complete understanding of individual customers, a company may need to use several models or views of its customers for accurate profiles.
Going back to the online music store, through these customers purchased the same CD, the customers’ past purchases show they should not receive the same recommendation. The 18-year-old has a history of buying older music as gifts. Knowing this, the site recommends a CD, based on past purchases for the 18-year-old, of the latest Metallica release. He purchases the recommended CD, even though that was not his intent, and the retailer has doubled this sale.
Though data analysis is a more exact route to personalization, business rules have a role in achieving effective personalization. After the recommendation engine uses customer data to make a suggestion, business rules are applied to ensure the recommendation makes sense.
In addition, business rules shape the predictive models used in data-driven personalization. Data analysis, working with business rules, enables a company to make targeted, personal recommendations to customers – the goal of effective personalization.
Deployment of analysis is the key element of effective personalization. The method above is useless unless the results of new data passing through a predictive model are seamlessly deployed from the analysis system to the recommendation engine. These computer systems need to talk with each other to deliver recommendations to customers and customer information back to the analysis system.
The following example illustrates effective personalization: Two customers visit a Web site that specializes in exotic teas. They are repeat customers who have both purchased green tea. Though business rules-based personalization would recommend both customers purchase a new Japanese green tea, the retailer has gathered data on each customer, including purchasing patterns, click-stream data, demographic information and opinion information including reasons for their interest in tea. The site will display different, personalized messages for each customer, even though it is, in fact, recommending the same item.
• Customer A is a woman who drinks tea for the taste. The Web site has been collecting information on this customer, and it recognizes her and knows that she recently purchased green tea and likes tea for taste. So, the site’s home page will provide a link to the new green tea and say, “Taste the purity of our Japanese Green Teas.”
• Customer B is a man who drinks tea for health benefits. Again, the Web site recognizes him and knows that he purchased green tea during his last visit and knows of his interest in health. Therefore, the home page will provide the same link, but his message will say, “Try our new Japanese Green Teas — the healthiest of the greens.”
These data-driven recommendations create a winning situation. For the customer, effective personalization gives a more focused, convenient and enjoyable shopping experience. For the online retailer, it drives customers to be return shoppers, which results in more profitable relationships with those customers — who may start to wonder whether you are, indeed, psychic.
• David Cody is senior manager, CRM/e-CRM solutions, at SPSS Inc., Chicago. Reach him at [email protected].