Techniques for Leveraging Atomic-Level Data

Many direct marketers need to be more creative in the ways that they leverage their customer data. Sure, just about everyone uses customer data to make general decisions about whom to promote. Typically, this is done with either rules-based logic such as RFM Cells, or more formal statistics-based predictive models. However, this is just scratching the surface.

The following two-step process provides a framework for you to become a more innovative data-driven marketer:

First, identify areas of your business that are suitable for focused, data-driven marketing. For example:

· High-potential customers, inquirers and prospects.
· Conversely, at-risk and lost customers.
· New ship-to customers such as gift recipients.
· Top-selling, or most profitable, products and services.

Then, take action that focuses on these targets, mixing and matching promotional vehicles and channels as appropriate.

The following are six examples of the many ways in which customer data can be creatively employed to help drive increased revenues and profits:

1: Identify commonality within customers.
This is also known as “customer clustering.” It is the creation of groups of customers for which, within any given group, every customer shares at least one thing in common. The groups are inherently interpretable because of this homogeneity. Therefore, they provide a robust foundation for highly-targeted specialty marketing programs.

2: Identify commonality within merchandise.
This is also known as “merchandise clustering.” It is the identification of merchandise that tends to get purchased together. This type of data mining can take two forms. The first is market basket analysis, which looks for merchandise groupings that take place within each transaction; for example, store visits for brick-and-mortar retail, and online visits for e-commerce. The second is product affinity analysis, which extends across multiple transactions.

3: Identify commonality within customer/merchandise hybrids.
Sometimes, breakthrough results can be achieved by crossing overlay demographics that describe customers – age, estimated income, marital status, presence of children and the like – with previously-purchased merchandise categories.

Homogenous hybrid groups, just as with their pure customer and merchandise-oriented brethren, assist with promotional content tailoring such as selective binding, differential ink jetting, and one-to-one email messaging. They also inform product placement decisions within e-commerce sites and brick-and-mortar stores, and provide the foundation for strategic line extensions.

The statistical tool of choice for creating homogenous groups, regardless of whether they are customer, merchandise or hybrid in nature, is either cluster or tree analysis, depending on the goals of the project.

4: Build merchandise-driven predictive models.
A frequent companion to merchandise clusters is merchandise-specific predictive models. Often, the homogenous merchandise groups are created, and then predictive models are constructed to help make the groups actionable. These models can take two forms. The first is up-sell models, which predict additional buying behavior within merchandise clusters for which the customer in question has made a purchase. The second is cross-sell, which predicts who will, for the first time, purchase merchandise within a given cluster.

5: Conduct data-driven operational CRM.
This is where relevant information is packaged for access at touch-points such as call centers and e-commerce sites, in order to support data-driven interactions. One example is the content-tailoring of call center scripts (“play books”) and Web pages to drive up-sell and cross-sell efforts using suggestive selling techniques. Another is best-customer identification during the resolution of service problems.

6: Implement marketing action/reaction systems (MARS).

These are automatic triggers that generate tailored promotions whenever customers display predefined behaviors. These behaviors can be positive or negative, such as:

· High-value customers with reduced recent activity, suggesting that they might be at-risk.
· Likewise, frequent previous purchasers of consumables whose recent volume has tapered off.
· Customers who purchased a product in the past that is approaching the end of its natural lifecycle.

MARS can also be employed to support rapid responses to competitive developments.

If pretty much all you’re using your customer data for is to make general decisions about whom to promote, then break out of your rut. Try creating some clusters, whether they happen to be customer or merchandise-oriented; or, hybrids of the two. Supplement your merchandise clusters with merchandise-driven predictive models. Explore embedding the results of your data-driven analytics into your operational CRM systems. Finally, implement marketing action/reaction systems to generate tailored promotions whenever customers display predefined behaviors. The reward will likely be substantial increases in your company’s revenue and profits.

Jim Wheaton ([email protected]) is a principal at Wheaton Group.