Ashley D. Su 
Companies rely on strong marketing strategies to increase sales, but the tools used to evaluate these strategies often provide misleading results, leaving managers unable to accurately measure how they can get the most out of their marketing investment.
“Companies really need to pay attention to the effectiveness of their marketing tools”
Thomas J. Steenburgh, an associate professor in Harvard Business School’s Marketing Unit, has developed a new analytical tool that more accurately measures the effectiveness of various marketing efforts. He created the model with Qiang Liu, assistant professor of marketing at Purdue University, and Sachin Gupta, Henrietta Johnson Louis Professor of Management and professor of marketing at Cornell University.
Steenburgh thinks the model could help brand managers determine the most effective marketing strategies to invest in.
“Companies really need to pay attention to the effectiveness of their marketing tools,” Steenburgh says. “They need to consider whether they’re creating new customers or simply driving customers away from competitors. That’s a fundamental question in the field, and this model helps measure that.”
The ideal mix
When planning marketing campaigns, brand managers have a wide range of weapons to rely on, including in-store merchandising, advertising, coupons and sweepstakes, trade promotions, prizes and the deployment of a direct sales force. The key is to create the right mix between them – the ideal brew needed to achieve sales and market share goals.
The trick is that each marketing effort affects consumer behavior in different ways and also elicits different types of responses from competitors. Some activities lead to an increase in demand for an entire product category. Take for example the “Got Milk” advertising campaign, which aims to increase demand for a product category, milk. On the other hand, an advertisement that emphasizes that a brand is better than a competitor’s brand aims to encourage consumers to switch products in a particular category.
If a company seeks to increase demand for a product category, the effort may not elicit much response from its competitors; after all, if the whole category grows, the rising tide lifts all the boats. But a competitor’s reaction is usually quite different when a company tries to increase market share, perhaps by offering price discounts. Since this strategy is considered more threatening, the competitor can be expected to retaliate with bias, often launching a campaign to win back many more customers than they lost.
“We know retaliation is happening and businesses are concerned about it,” Steenburgh said. “But no one benefits when both companies fight back. One effort just compensates for the other.”
Measuring the different effects of these marketing strategies can help brand managers make the right decisions about which strategies to use in their marketing mix. Steenburgh, Liu, and Gupta argue that tools that have been used in the past to analyze the effectiveness of different marketing activities, called discrete choice models, can skew results and mislead brand managers.
Traditional discrete choice models — logit, nested logit and probit, for example — are flawed because they make all marketing activities appear to produce the same results, the researchers say. In reality, the differences between different marketing instruments are often significant. The cause of these erroneous results comes from what is called the Invariant Proportion of Substitution (IPS) property, which implies that the proportion of demand generated by taking business away from a competitor is the same regardless of the marketing activity used. .
“These models are used all the time in academics,” Steenburgh says. “There have been discussions at conferences where there seems to be an understanding that these models are too restrictive.”
Widen view
So the professors created a new discrete choice model called Flexible Substitution Logit (FSL), described in their working paper The Flexible Substitution Logit: Uncovering Category Expansion and Share Impacts of Marketing Instruments. The model relaxes the IPS property and allows a greater variety of outcomes to be analyzed when studying the effects of different marketing instruments. In doing so, “FSL allows a greater variety of individual-level choice behaviors to be retrieved from the data,” the researchers say.
The team tested the new model by examining the marketing of prescription drugs, namely statins, used to lower cholesterol levels in people at risk for cardiovascular disease. Using data from 2002 to 2004, they studied the three main ways Pfizer, Merk, Bristol-Myers Squibb and AstraZeneca market these drugs: “retail”, in which pharmaceutical company representatives personally visit to doctors to sell the drug; at meetings and professional events (M&E) sponsored by pharmaceutical companies; and using direct-to-consumer advertising (DTCA).
First, they used the complex mathematical formulas of traditional models to study different marketing strategies used by pharmaceutical companies. They found that the IPS property created counterintuitive estimates of demand gains attributable to these marketing investments. Although the researchers logically expected the details to generate greater demand for the products than direct-to-consumer advertising or meetings and events, traditional models wouldn’t allow them to find out because of the SPI.
When they applied their FSL model, however, the results provided much more detail about the potential effects of different marketing investments. For example, the model predicted that DTCA and M&E sales gains would come primarily from category expansion (87.4% and 70.2%, respectively), while retail gains would come at the expense of competing drugs (84%). In contrast, the random coefficient logit model predicted that gains in PDCA, M&E, and detail would largely come from competing drugs.
“The FSL model is very useful if you want to predict consumer demand,” says Steenburgh. “This template gives you a better way to do it.”
Feature in Recovery
With results that provide better insight into how different marketing instruments work, brand managers can now decide how best to invest their marketing dollars. For example, if a brand manager is concerned about retaliation from competitors, the best decision may be to limit investment in retail and instead focus on direct-to-consumer advertising or sponsoring meetings and of events, both of which are more likely to develop the category.
Steenburgh notes that future research is needed to find alternative models that overcome IPS, and he hopes the FSL model will be applied in further studies that examine the effectiveness of marketing instruments.
“It would be interesting to apply the FSL model in many other situations to see which ones make the cake bigger and which ones threaten other actions,” he says.
