While there are a variety of approaches to predict what an existing customer will purchase next, one very effective approach is to calculate the product adoption propensities within a single model, based on the idea that a customer is choosing between various products. For example, the cash a customer brings to a bank might be placed into a savings account, CD, or investment account, but a particular customer is likely to open only one such account at any time. Similarly, in retail, consumers make choices between product categories in any particular shopping trip, although the "market basket" may contain products from more than one category.
The output of such a model is an array of customer-level scores where the columns represent product categories (for example, electronics, books, appliances, furnishings, and apparel). If we ordered customers in descending order by any product column, then we could use this model as a traditional targeting model. If we look by row, then we can determine the best product(s) to offer any given customer.
No matter what method we use, the value of such a model cannot be realized until it is used to deliver relevant offers and messages. Unless you are one of the rare marketers with the luxury of engaging customers through a single channel, you will need to deploy this customer intelligence across channels. Unfortunately, many marketers must still rely on a hodge-podge of single-channel platforms to deliver messages through various channels. To get around this problem, consider the following steps:
Establish policies and procedures to ensure that customers receive consistent treatment regardless of channel. It makes no sense to communicate one offer through a web site, sales rep, or mail piece, only to send out a contradictory message through email.
Use the cross-selling model to develop a library of customer-specific offers. For example, if the model recommends offering a credit card, take the extra step to formulate that into an offer for that customer. Now, instead of a model score, you have a more useful piece of information that can be deployed.
Create a mechanism for pushing these pre-built customer-offer combinations out to touchpoint solutions. For real-time systems, such as web sites, kiosks, and call centers, web services may facilitate more rapid integration.
Lastly, if you find it increasingly difficult to manage multiple channels – whether or not you are trying to cross-sell – then consider investing in a real cross-channel platform.
David King is EVP of Fulcrum, a leader in advanced analytics, technology and multichannel program solutions for marketing. He can be reached at firstname.lastname@example.org.