Predictive modeling, also known as list scoring, is an underused tool that can boost outbound response rates for telemarketers working off large lists with significant customer data.
Lack of awareness, technical expertise or both has thus far prevented many potential users from tapping into the advantages of predictive modeling. But if attendance at a presentation about the topic at the recent Chicago Direct Marketing Days is any indication, teleservices agencies want to find out more about it.
“Over the last five years, [clients have begun] to see benefits of it,” said Laura Hansen, vice president of sales and marketing for Direct Response Corp. (DRC), Glenview, IL, who made the presentation. “When using outbound calling, it’s got to be productive and profitable. Modeling is a win-win because your productivity is increased as is openness to your message.”
Modeling can be applied to telemarketing in much the same way it has been successfully applied to direct mail. Ed Vesely, executive vice president of teleservices for the Signature Group, Schaumburg, IL, points out, however, that scoring models for mail differ from models for telemarketing and are specific to every product and list.
The modeling or scoring process involves building a model that draws on known customer data, such as past buying behavior to predict future behavior. Buying patterns for example, can be overlaid with various demographic attributes to identify characteristics common to all buyers. Effective models are able to predict which prospects on a list are the most likely to respond to offers and steer outbound calling to that highly targeted audience.
“By targeting the top 40 percent of the prospects, the resulting response rate will increase by 40 percent,” Hansen said in detailing the results of a case study on cable TV buyers.
Modeling can also be used to eliminate segments least likely to respond. Hansen said such a weeding out process will increase response rates by 23 percent. The net effect is lower costs per sale and a lower number of calls needed to meet sales goals.
“For a client with any amount of data available on their customers and a list of sufficient size, this system definitely works,” said Tom Salutz, president of Results Management, Chicago, who has been constructing predictive models for telemarketing since 1988. “If you have only a little information or a small list, a model won’t help.”
Modeling is also not applicable when outbound calling is used for lead generation or to establish an initial contact such as in fundraising.
Convincing skeptical clients to give models a try has been a stumbling block to its growth in telemarketing. Those who have used models in direct mail campaigns are an easier sell. Hansen said DRC has provided modeling to its cable TV clients on a cost-per-sale basis.
“Clients are scared to work just the top 40 percent,” Hansen said. “They’re afraid they will miss a lot from the bottom 60 percent.”
For a medium where getting people to stay on the phone is critical, modeling can screen out those overmarketed suspects who are easily angered by intrusive calls. Calling only prospects willing to listen–existing cable customers in the case of DRC–makes life easier on both the customer and the call center representative.
Another stumbling block has been ignorance or lack of access to the technology. Builders of predictive models don’t aggressively market their services, so agencies and outsourcers may not know that models even exist.
Clients of the Signature Group have no such concerns because list scoring is done internally and has been a standard practice since 1985. Nevertheless, model building requires a statistical expertise that most service bureaus don’t have. Vesely said most service bureaus can refer their clients to a company that does develop models.
Despite these problems, the familiarity and use of models is increasing.
“It’s obviously much better understood and appreciated on the direct mail side, but most of the major players on the telephone side are aware of it,” Vesely said. “The desire to use [scoring] is determined by who is paying.”
