Segmentation Functions

The central purpose of a high- end campaign manager is to handle large, complex segmentations. This requires several features.

Nested segment definitions. Typical high-end marketing campaigns are built by segmenting a file on several dimensions. These might be the classic recency, frequency and monetary value of catalog marketers, account types and deposit levels in financial services or usage and service categories for telecommunications. Any given campaign could use three to ten dimensions, each with multiple categories. These categories may not be mutually exclusive; for example, a customer who purchased several different products would fall into several categories. The high-end campaign manager needs an efficient way to define a separate segment for each combination of dimension values. It also must let the marketer establish a hierarchy among the segments to determine the assignment of customers who fall into multiple categories. Also, it must allow marketers to combine categories when quantities become too small to treat separately.

This is usually handled through nested segment definitions. That is, the user defines a set of segments using values from a single dimension. The user then splits each of these segments into more segments based on values for a second dimension. These, in turn, are split using values from yet another dimension, and so on, until the final combination of values to define each promotion segment is identified. Customers who qualify for one segment are usually excluded from subsequent segments in the same dimension; therefore, if there is a split of shoe buyers followed by dress buyers, all shoe buyers are automatically excluded from the dress-buyer segment. Occasionally, this sort of duplication is desired; therefore, most high-end systems give users an option to allow it.

Complex splitting rules. In most cases, segments are split on very simple criteria such as membership in a specific group (e.g., gold customer) or a mathematical relationship (e.g., account balance greater than $10,000). Categories such as "gold customer" may themselves be defined through complex calculations performed outside the campaign management system and indicated by a code on the customer record. Sometimes the segmentation needs to generate its own complex criteria. One common requirement is related to counting (e.g., customers with more than five electronics purchases in the last six months). If the same count is used frequently, it can be precalculated and stored on the customer record. However, often a marketer will need a particular count – perhaps involving a specific time range or set of products ­– that was not anticipated. In these cases, the campaign manager must provide the ability for end users to specify the necessary calculation as part of the segment definition.

Ad hoc counts are just one example of a complex splitting rule. Other types that arise with some regularity include set-based calculations (purchased any three of a group of five products) and not relationships (never purchased shoes). These types of queries are difficult to write in SQL. Less frequently needed, but also difficult in SQL, are queries involving aggregates (e.g., number of balance inquiries more than 30 percent above the average for all bronze customers) and ratios (e.g., number of phone calls in the past month 20 percent lower than number of calls in the preceding month). A high-end campaign manager provides ways to perform these sorts of queries without precalculations or Ph.D.-level SQL code, often through prebuilt templates or special query functions.

Sampling and quantity limits. Serious direct marketing programs –­ the kind that require high-end campaign managers –­ are built around champion-challenger testing. This involves simultaneous execution of alternate marketing approaches and comparing the results. The approaches must run against statistically identical audiences so that any differences in results can be attributed solely to the different approaches. This means the campaign manager must be able to split segments into subsegments through random sampling, which ensures that the subsegments are comparable. True random sampling is sometimes not available, in which case marketers are usually willing to accept "Nth" samples, in which the system takes every second, third, fourth or whatever (Nth) record for the sample. Typically the user specifies a target quantity for the sample segment(s) and creates an additional segment to hold any names that are left over. In some cases, though, users specify a percentage allocation and let the system assign to each segment however many records this produces. High-end systems allow both approaches.

Although champion-challenger testing is the main reason that high-end campaign managers require sampling functions, there are also other reasons to split a segment based on percentages or a maximum quantity. These include requirements to allocate promotions among different products in specified proportions or to limit the quantity to the number of mail kits available. Where quantity limits are involved, the user may want a ranked selection –­ to take the best prospects –­ rather than a random selection. A high-end campaign manager can do the ranking on any user-specified numeric variable or according to a user-specified formula.

Automated value splits. Sometimes a campaign needs separate segments for each value of a particular attribute. Common examples are state codes (important in insurance marketing, where each state gets a different offer due to regulatory differences), bank branches and retail locations. It's possible to code each of these splits manually; however, it is much easier to have the system set them up automatically. Marketers may need to combine automated splits with quantity limits and ranking to handle selections such as "top 100 customers per store."

Names within households. Campaign segments are often defined at the household or business level, but still sent to individuals. A high-end campaign manager can select the best individual within a larger entity, while still selecting the entity based on its own attributes. In addition, particularly in business marketing, the system should allow users to specify a maximum number of names per entity. The names themselves should be chosen based on user-defined ranking criteria, which ideally include a priority scheme for nonnumerical values such as business titles. Some systems can generate titles when actual names are not available.

A Note on Multistep Campaigns

Classic high-end marketing campaigns involve a single promotion that generates direct purchases or leads that are turned over to other channels for follow-up. Thus, high-end campaign managers do not necessarily have to support sequences of promotions within a single campaign. However, because such sequences are an important objective for many marketers who purchase such systems, many systems do allow them. Requirements for multistep campaigns will be discussed later in this series.

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access