The first step toward customer relationship management is installing front-office systems that access a shared history of customer interactions.
This ensures that a customer will be treated consistently whether she contacts the organization in person, by telephone or electronically. In fact, most of the time, money and attention devoted to CRM is spent on systems for call centers, sales forces, Web servers and field locations. The heavyweights of the CRM industry are the plumbers — big system integrators and front-office software vendors who lay down pipes without worrying too much about what goes inside them.
Still, companies will eventually turn their attention from making interactions consistent to making them consistent and appropriate. This would seem to be a job for campaign management software, which is, of course, designed to get the right message to the right customer at the right time. But conventional campaign management systems are built to generate lists of promotion targets, which means they focus on segmenting large groups of records. Real-time interactions pose different technical challenges — moving data efficiently between the front office and marketing systems, and responding appropriately as situations evolve with each new input. So new solutions are needed.
RightPoint Real-Time Marketing Suite (RightPoint Software, 650/287-2000, www.rightpoint.com) offers an impressively comprehensive approach to real-time interaction management. RightPoint itself was originally founded in 1994 as DataMind, a developer of predictive modeling software. But it has since renamed and reinvented itself to focus on this new market. Its product suite offers a half-dozen interrelated components.
Several of these help connect front-line events with the marketing decisions. A module called Campaign Workshop lets administrators define the data elements available for decision making, including information captured during the event by the front-line system and data read from a warehouse or operational system.
Technical details, such as the actual query used to locate a specific customer record, are hidden inside an intermediate object called a customer profile. This means that marketers simply see a list of available data elements with descriptive information.
Data from the front-line systems is transmitted by RightPoint components placed inside those systems. These components, Call Point Manager and Web Point Manager, also receive replies from the central RightPoint systems. The replies may be as simple as a Web page address or a message inside a pop-up window or as complicated as a trigger that executes functions within the front-line system itself. The call center module includes a basic script writer, although dynamic branching would require treating each branch as a separate campaign. Information moving in both directions is encapsulated in Component Object Model objects, which streamlines new connections. Modules for interactive voice response, fax and e-mail are scheduled for release in May.
Another set of modules determines the replies themselves. Control Center, accessed through a Web browser, lets marketers define campaigns and offers within campaigns. Each campaign has a set of rules that determine whether it's considered when choosing the reply to a particular event. If the campaign is activated, each offer within it is evaluated against its own set of rules.
When more than one offer qualifies, another set of rules chooses among them. Having several layers of rules simplifies marketing program design and reduces the amount of processing associated with any one event.
The rules associated with campaign activation usually involve simple conditions such as a date range or system activity level (for example, to skip the cross sell messages when the call center is very busy). The offer qualification rules include both simple conditions and analytical results such as a model score indicating the likelihood of response. Arbitration among multiple offers may be based on response likelihood or expected benefit, which is the response likelihood multiplied by a user-specified benefit value. The benefit value could be a specific revenue or profit amount, or — to incorporate nonfinancial considerations — a generic ranking from 1 to 10. While response likelihood and expected benefit are prebuilt arbitration methods, users could also create their own.
RightPoint offers several methods of calculating the response likelihood.
The DataCruncher module, chief remnant of the original DataMind, lets skilled users create models using either the vendor's Agent Network Technology or conventional neural networks. These models are registered in Campaign Workshop, so the score appears as an element of the customer profile.
Alternately, attributes can execute third-party model algorithms or call external scoring systems. Finally, RightPoint's Campaign Server — the module that actually receives input, executes campaigns, and returns replies — has a built-in Real-time Miner. This automatically generates response models and then updates them continuously as offers are made and accepted. For all types of models, scores are calculated only when they are needed to resolve a particular event.
The system includes a variety of refinements needed for effective marketing. These include arbitration rules to generate random tests of alternative offers; deployment of different campaigns to different user groups, such as sets of call center operators; robust treatment of missing data; and automatic logging of failures to return a reply when one is expected.
RightPoint allows the same offer to be used in multiple campaigns, provides standard reports on acceptance by campaign and offer, and can capture acceptance even when it occurs after from the original event. The system currently captures statistics on model performance, although these will not be presented in standard reports until later this year. A more significant limitation, also being addressed, is that Real-time Miner does not automatically weigh recent transactions more heavily than older ones–making the system slow to adjust to changes in customer behavior. Nor can users enter forecasts for campaign results, so the system cannot highlight performance above or below
RightPoint runs on Windows NT and Unix servers. It can run on multiple servers to handle high volumes. Campaign Workshop connects directly to Oracle and Informix databases and via ODBC to others. Pricing depends on the modules, number of servers and number of call center stations. A minimum installation would be about $200,000. The system has more than ten installations.