LifeTime System Manages Behavior in Real Time

“Right message to the right customer at the right time” has been the mantra of direct marketing gurus since the industry began. What changes is technology: New media become available to deliver the messages and, in turn, place new requirements on the systems that make the selections.

The needs of today’s newest medium, the Internet, include immediate response as transactions occur, accommodating a near-infinite variety of situations and managing the flood of detail generated by online interactions. Not surprisingly, a new set of products has appeared to meet these requirements. Contenders include RightPoint (recently acquired by E.piphany), Trivida, Black Pearl and Harte-Hanks Allink Agent.

LifeTime (Verbind Inc. also is built to manage real-time interactions. Like other products in this group, LifeTime primarily addresses the mechanical issues of executing business rules that have been designed offline by clever marketers. LifeTime uses what the vendor calls behavior maps to address the three issues of data volume, situation variety and real-time response.

In LifeTime’s terms, a behavior map refers to a set of categories, or locations, that customers can fall into. Each category is defined by a combination of customer attributes such as traditional demographics and observed behaviors. These behaviors can include specific transactions or patterns of transactions such as a weekly bank deposit.

A typical map will have from 20 to several hundred locations – a small enough number for humans to manage effectively, even though there will be subtle differences among customers within the same location. To avoid oversimplification, the system allows several maps to be active simultaneously. For example, one map might deal with customer value while another measures likelihood of attrition. In large institutions, different customer groups might be assigned to different managers, each of whom would be limited by security to changing and using their own set of maps.

Using maps helps solve the data volume issue in several ways. The location definitions specify which transactions are important, so the system knows it can ignore the others. Once a transaction has been processed and the customer’s location is adjusted accordingly, LifeTime generally can discard even the significant transactions, since the customer’s current and prior locations provide most of the information needed to select marketing messages. And since a location can be based on a pattern of events, the system can scan for an expected event and report it has not occurred without reviewing the entire transaction history in detail.

Although maps provide the basic framework for analyzing customer behavior, they are not enough to select marketing messages directly. Instead, LifeTime bases messages on the audience a customer belongs to. Audiences can be defined in the context of the maps, by specifying some combination of a customer’s current location, time in a location, movement among locations, specific events and cumulative number of events such as customer service requests. If several maps are active, an audience definition can consider the customer’s location in all of them. Definitions also can include filters that set priorities among messages and limit the total number of messages a customer receives. Messages themselves can include keycodes, Nth selects and random splits for traditional champion/challenger testing. Groups of messages can be assembled into campaigns, which are assigned start dates, end dates and execution intervals for promotions.

LifeTime can work with data from both conventional and online data sources. In either case, the system accepts a stream of transactions from the source system and selects those that have been identified as relevant. These are processed through the behavior maps and definitions, and the results are stored in updated customer objects using a highly compressed format. Each object contains the data, messages and decision rules related to a particular customer.

Specialized technology developed by the vendor lets the system scan several million transactions per hour in this fashion. The system’s ability to handle high volumes lets it accept all transactions generated by source systems rather than asking the client’s IT group to develop special extracts. This simplifies implementation.

In addition to conserving storage space and speeding updates, the customer objects help LifeTime to support real-time interactions. When the system is set up to operate this way, a component of LifeTime is attached to the online system itself – say, the Web server – where it calls for a current customer object when one is needed. The system combines information in the object with new data as it is captured by the online system and produces whatever messages are appropriate. Because the object is self-contained, only one transmission from LifeTime is required.

LifeTime was initially deployed in 1998 and officially released in February 1999. Initial implementation involves a one- to two-month consulting engagement during which the vendor analyzes one year of transaction history and designs an initial set of behavior maps. Clients can add audiences and messages and modify the maps themselves over time. In addition to tactical reports on marketing results, the system can produce migration reports that show the movement of customers among different map locations, illustrating the long-term results of marketing strategies. These features help users refine their behavior maps and definitions, although LifeTime does not suggest changes based on any sort of automated analysis.

Pricing of LifeTime reflects the data volume and complexity of a customer’s requirements. Initial installation costs $75,000 to $400,000, while monthly fees range from $10,000 to $150,000. The system has four operational clients with several additional installations under way. Users are in financial services, e-commerce and other industries where frequent customer interactions provide opportunities to react quickly to changes in customer behavior.

Related Posts