U.S. companies spent a staggering $285 billion on advertising in 2006, according to Advertising Age. That's a lot of dollars and expectations being handed to advertising managers to generate returns.
But do marketing managers allocate their resources wisely? Are traditional methods of resource allocation such as heuristics, decision rules, and "bottom up" approaches still effective in today's complex marketing environment?
To help practicioners answer this question, Harvard Business School professors Sunil Gupta and Tom Steenburgh surveyed the academic literature as well as practical examples to create a framework to help managers think about their allocation challenges. The result was published in a recent paper, "Allocating Marketing Resources".
Brand managers can be confronted with a complex, daunting task, the authors note. "Marketing resource allocation decisions need to be made at several levels—across countries, across products, across marketing mix elements, across different vehicles within a marketing mix element (e.g., TV versus Internet for advertising). Each decision requires some specific considerations."
The good news is that an increasing availability of data, combined with more sophisticated methods of analyzing it, provide managers with powerful tools to help them isolate the effects of various marketing instruments. The framework developed by Gupta and Steenburgh helps managers think through the process, in part by looking at examples of what has worked for firms in the past, and also by reviewing more theoretical approaches developed by academia.
The framework works in two parts. In part one, the manager estimates a model of demand. In part two, the estimates are plugged into an optimization model. The results will help managers decide which existing customers to target, which new customers to acquire, and how to balance resources between promotions and advertising, allocate promotional dollars, and determine the effectiveness of word-of-mouth communications.
We asked the researchers to offer more details on their research.
Sean Silverthorne: What was the genesis of the research that ended up as "Allocating Marketing Resources"? What was the problem that you were trying to answer?
Sunil Gupta and Thomas Steenburgh: Our goal in writing this chapter was to ask how the academic literature might be useful to managers who have to make marketing allocation decisions.
Many interesting and useful techniques have been developed over the past 20 years that might be brought to bear in addressing these problems, and we thought it would be great to step back and develop a framework that helps managers think about how these methods are useful and where they might be applied.
We want to make common academic techniques more accessible to practitioners and to encourage their use in solving real business problems.
Q: Many managers continue to use traditional methods for allocating marketing resources, such as heuristics and decision rules. What is happening that makes these tools less effective today?
A: Companies have always spent a great deal of money on marketing, but data have not always been available to determine whether this money has been spent wisely. Companies may have been able to track broad measures of brand equity—for example, how well consumers were able to recall a brand name after commercials have been run—but not much more. The effectiveness of marketing investments was rarely directly tied back to sales.
This has completely changed with the revolution in information technology. Now, even companies that have millions of customers are able to track how marketing investments affect individual customer's behavior. Every aspect of behavior is being monitored—a customer's first encounter with the firm, the initial and subsequent purchases, and the post-purchase service experiences. As a result of this information, managers are being held to higher standards when it comes to justifying customer investments.
We foresee the need for marketing professionals to develop even greater analytical skill as the field continues to evolve. This should be a very positive development because other employees in the firm will be able to more clearly understand when and why customer investments are necessary.
Q: Describe the framework that you have developed for managers. What are its goals? Is it practical?
A: The framework is both simple and practical. It gives managers a way to conceptualize an approach to making allocation decisions—in essence, a way to organize their thoughts.
Basically, all marketing allocation problems need to be addressed in two steps.
In the first step, an analysis is undertaken to predict how different marketing actions will affect consumer behavior. This can be accomplished in three basic ways:
- Decision calculus: Knowledgeable people use their judgment in a systematic way to predict how consumers will respond to actions.
- Experiments: The company trials different marketing actions, either in the field or in a lab, on one group of customers and compares how that group responds against a control group of customers.
- Econometric analysis of historical data: Historical data are analyzed to determine how customers have responded to different marketing actions in the past. Predictions can then be made as to how either they or a new set of customers will respond in the future.
In the second step, the manager reflects on the demand analysis that was performed and asks which of the potential actions that might be taken would be best for the firm. This too can be accomplished in three basic ways:
- Descriptive approach: Summary measures, such as the elasticity of demand, are calculated so that directional recommendations can be made.
- 'What if' analysis or simulations: A small number of marketing plans are considered by simulating their effects on sales and profit. The best plan can then be chosen from this set.
- Formal optimization: An optimization algorithm is written to determine which action or actions maximize a specific objective.
In the chapter, we discuss the strengths and limitations of each of these methods and illustrate their use with case studies and applications.
Q: Give an example of how a marketing manager might use your work to create a marketing allocation plan. Where would the starting point be?
A: It is well known that some marketing actions, such as supermarket promotions, deliver short-run results, whereas others, such as television advertising, deliver long-run results. Yet, firms have a difficult time deciding how to allocate resources between these two marketing actions.
Jedidi, Mela, and Gupta (1999) addressed this problem by developing an econometric model. They were able to quantify the short- and long-run effects of promotions and advertising for a non-food consumer packaged good using household purchase data. They found that promotion and advertising had long-run consequences on consumers' buying behavior in two ways: by influencing the brand equity and by affecting the consumers' price sensitivity.
Jedidi and colleagues then simulated how different promotion and advertising levels affected short- and long-term demand for this product. Interestingly, they were able to use the model to predict how competing firms might react to different marketing scenarios, which led to better estimates of the real changes in demand that would result from a new marketing plan.
In the end, the company, a well-known consumer packaged goods firm, used this study to change the way it marketed this product.
Q: I found one of the most powerful take-aways from your work was just seeing in one place how many exciting approaches have been developed over the years, how the science of marketing has advanced. Can you talk a little about the impact of these advances?
A: One of the biggest impacts of this work has been the trend toward micro-targeting. A company may have millions of individual customers, but now each one receives a customized message for a customized product at a customized price. Companies are able to create sustainable competitive advantages by developing a deep understanding of what their customers want and how they behave. It can be very difficult for the competition to pick customers off because they do not have the same level of understanding.
One of us (Thomas Steenburgh) has just written a case on a company called ScriptLogic that runs its business in this new fashion. They describe their business model as "click-to-cash." ScriptLogic has developed a system that tracks which advertisement has driven an individual prospect to its Web site. They then track each prospect's behavior from the initial visit to the first download to the first purchase and beyond. Not only has ScriptLogic been able to invest in marketing actions that deliver the best results, they have been able to use their data to strike better deals with advertisers.
Q: Your work mentions Harrah's Entertainment as an exemplar in its ability to improve performance by understanding and predicting customer behavior. Can you illustrate some of the things that Harrah's does in this area?
A: Harrah's does a great job of micro-targeting. They constantly run new campaigns to see what actions change the behavior of individual customers, which would not be possible without a team of "propeller heads," as (CEO) Gary Loveman likes to call them.
The old way to run a business was for companies to spend money on their best customers. Harrah's embodies the new way of doing business: they spend money on customers when it makes a difference in how they behave.
Q: What are you working on next?
Sunil Gupta: Resource allocation becomes much more complicated when consumers interact with each other. I am spending much of my time thinking about how social networks change the way we think about marketing allocation decisions. I am working on papers about "The Value of a Free Customer" and "The Value of a Customer in a Social Networked World."
Thomas Steenburgh: I am spending time thinking of a few different things. I have a working paper with a few graduates from our DBA program and John Deighton on how brick and mortar stores affect direct channel sales. I also have been rethinking what can be learned from some old models of individual choice behavior. This is a long-run project, but I think that my work will lead to new models that make better predictions in micro-targeting applications.