JUNE 2004


• • • • • • • • • • • • • • • • • • • • • • • • • •

Marketing Prescription Drugs

How Effective Are Direct Sales Calls?

Research by Puneet Manchanda, Peter E. Rossi, and Pradeep K. Chintagunta

New research provides a way for pharmaceutical companies to measure the effectiveness of direct sales calls on individual physicians.

A marketing manager's toolkit is composed of the four P's: price, promotion (advertising and direct selling), place (distribution), and product. Taken together, these four P's-the "marketing mix variables"-compose the marketing strategy that companies use to market their product to different customer segments. Each variable acts as a lever that can be used to increase sales. Managers make "optimal" marketing mix decisions based on the extent to which sales increase in response to changes in each of the marketing mix activities.

Academics and marketing managers have long been interested in understanding how the four P's work. Historically, researchers and practitioners have relied upon statistical models, called "sales response models," to quantify the effect of these variables and come up with the best mix. These models are required to estimate the sensitivity of sales to a change in a marketing mix variable-e.g., how much do sales of a product increase if the price is reduced by 10 percent?

In a new study, "Response Modeling with Non-Random Marketing Mix Variables," University of Chicago Graduate School of Business professors Puneet Manchanda, Peter E. Rossi, and Pradeep K. Chintagunta go beyond the standard sales response model to determine the effect of different marketing mix variables. Their model looks at the effect of the marketing mix instruments on individual customers rather than at the aggregate level. They apply their modeling approach to the problem of gauging the effectiveness of direct sales calls to induce the greater prescribing of drugs by individual physicians.

The marketing mix variables for pharmaceutical companies include: direct sales calls; advertising directly to consumers; advertising in medical journals; and advertising at various physician meetings.

Of these variables, pharmaceutical companies spend the most money on direct sales calls, with more than $6 billion spent in 2000. In a typical scenario, a physician attends brief individual meetings with more than a dozen salespeople from different pharmaceutical companies on a given day, all of whom will bring samples and tout their company's latest drugs.

For pharmaceutical companies, it is important to figure out the increase in the number of new prescriptions written for the particular drug as a result of a direct sales call to that physician.

"Pharmaceutical companies spend nearly three times as much money on direct sales calls than on advertising to consumers," says Chintagunta. "That's why it is important to measure how well pharmaceutical sales force efforts are being allocated."

The authors find that pharmaceutical companies are not efficiently allocating sales force efforts for at least 50 percent of the physicians in the study. High-volume physicians, those who write more prescriptions, receive far more direct sales calls than low-volume physicians, seemingly without regard to the physician's responsiveness to these visits. It appears that unresponsive, but high-volume, physicians receive the most direct sales calls.

The Physician as Individual

Manchanda, Rossi, and Chintagunta used data from a major U.S. pharmaceutical firm that provided information on physician prescription writing behavior and sales force efforts for a drug in a mature product category. The "mature" classification indicates that the category is not expanding, and therefore total sales levels for the drug are not trending up or down. The drug is part of an important therapeutic category with an estimated 19 million patients in the United States.

The data included a monthly record of how many prescriptions each physician wrote for the drug from June 1999 to June 2001. From internal firm records, the authors accounted for the number of direct sales calls made to each physician per month, as well as the number of free samples distributed to each physician.

From a total sample of more than 112,000 physicians, the authors chose a random sample of 1,000 physicians for the study. On average, physicians in this sample wrote five new prescriptions for the drug per month, and received two direct sales calls and six drug samples per month.

In a typical sales response model, the aim is to find out how changes in a marketing mix variable affect sales. In such analyses, the marketing mix variable is taken as a "given." However, in many industries, the marketing mix variable is set strategically based on prior information about the company's consumers.

In this study, the authors were told by the pharmaceutical company that they make more calls to physicians who write more prescriptions. The pharmaceutical company producing the drug does not set direct sales call levels randomly, and instead contracts with a consulting firm to help optimize the allocation of their national sales force. The consulting firm recognizes that targets for direct sales calls should ultimately be set at the physician level. The consulting firm and the pharmaceutical firm managers cooperate to set direct sales call targets on an annual basis.

Allocating direct sales calls based on the volume of prescriptions written assumes that the marginal effect of sales calls on the probability of prescribing the company's drug is the same for all physicians.

"Pharmaceutical companies have been making direct sales calls to these physicians for a long time, and have some notion as to which physicians are responsive," says Chintagunta. "As a consequence, the way they set the sales force effort must be a function of their knowledge of this responsiveness, which adds another layer of complexity to the response model. You have to take this knowledge into account to get the appropriate estimate of the effect of direct sales calls on prescription writing."

Sales Response Modeling

Marketing is increasingly focused on the individual consumer, and as a result there has been an increase in the use of a statistical method referred to as Bayesian hierarchical modeling, which helps predict individual consumer behavior in response to changes in marketing strategies.

"For a particular physician, what is the impact of direct sales calls on his or her prescription behavior?" asks Chintagunta. "That level of understanding is made possible by Bayesian hierarchical methods which help researchers drill down to individual analyses."

The model developed by the authors is based on two equations: 1) a response equation, which calculates the level of prescription writing as driven by direct sales calls; and 2) a direct sales call equation, which builds on the parameters of the first equation and uses the base level of prescriptions and the responsiveness of the physician to sales force efforts as a foundation for evaluating the direct sales call level itself. The use of responsiveness to determine the level of calls is a modeling innovation proposed by the authors. Thus, the authors' conditional response model allows them to differentiate between physicians who are big or small prescribers as well as those who are or are not responsive to direct sales calls.

The authors find that the least responsive physicians receive almost twice as many direct sales calls as the most responsive.

"Physicians who are actually very responsive to sales calls are being visited far less frequently than they should be," says Chintagunta. "If the company transferred their sales force effort away from the least responsive physicians to the most responsive physicians, the company would likely see higher sales for the drug."

The authors find an excessive amount of sales calls were made to the top 20 percent of prescription writers regardless of their responsiveness to direct sales calls. However, the authors do note that the most frequently visited physicians also may receive many sales calls from competing drug companies, which may lower their responsiveness to the average direct sales call. Managers may not have access to data on competing sales calls when making their sales force allocation decisions.

To increase the number of prescriptions written for the drug, direct sales calls should be focused on physicians who are most responsive to the visits, rather than those who write the most prescriptions.

Optimal allocation means making sure those two elements, direct sales calls and responsiveness, line up. The authors find that the two elements did not line up for about 50 percent of the physicians that the firm targeted. This implies that there is significant potential for improvement in the sales efforts of pharmaceutical firms.

Chintagunta suggests that the most efficient allocation of sales force efforts would be as follows: "Take two physicians: one is very responsive to direct sales calls, the other is not responsive at all. The company should send its salesperson to the physician who actually responds to these direct sales calls by writing more prescriptions. Otherwise, the company will be wasting its resources."

Applications for Other Industries

While physician data is available to most pharmaceutical firms, it is difficult to implement these models because of the technology and skill required to use advanced statistics to calculate individual level response. As is the case with most research, there is likely to be a time lag between new findings and industry practice.

However, as technology advances, it will become easier for marketing managers to perform these calculations and implement these response models.

"If a company can understand at the individual level how consumers respond to a promotion, then they can use that information to tailor their marketing programs," says Chintagunta.

"Furthermore," adds Manchanda, "our approach can be applied to any industry where there is the potential to strategically set marketing mix variables. For example, advertising in different markets may be set based on how big the market is and how responsive each market is to the advertising. Our approach can be used to determine whether an advertising expenditure is the optimal amount and how the allocation of resources can be improved."

Most importantly, these methods can create a better understanding of the efficiency and effectiveness of marketing activities, which can help companies obtain a higher return on their marketing investment.

 

• • • • • • • • • • • • • • • • • • • • • • • • • •

Puneet Manchanda is associate professor of marketing at the University of Chicago Graduate School of Business. Peter E. Rossi is Joseph T. Lewis Professor of Marketing and Statistics at the University of Chicago Graduate School of Business. Pradeep K. Chintagunta is Robert Law Professor of Marketing at the University of Chicago Graduate School of Business.

>>View the study: Response Modeling with Non-Random Marketing Mix Variables