John Little, the Founding Father of Marketing Science, on how the discipline has evolved
John Little, Institute Professor of Management Science at MIT’s Sloan School of Management, is widely regarded as the founding father of marketing science. While he began his career in the field of operations research almost five decades ago, he eventually branched out into marketing and became known for his research on models of individual choice behavior, marketing mix models for consumer packaged goods, and adaptive control of promotional spending. In operations research he is best known for his proof of the queuing formula, commonly known as Little’s Law.
Little has been at the forefront of many areas-even before the internet became popular, he had already created and implemented the idea of online marketing models. He was one of the pioneers in the use of scanner data, and predicting future buying behavior. He started companies such as Management Decisions Systems and Kana Software. The annual John D.C. Little Award is awarded by INFORMS in his name.
Even at the age of 84, Little is sprightly as ever. He works just as hard, and still goes for a run each morning. He showed no signs of jet lag when he flew down from Boston to Beijing where he spoke about his pet subject-marketing science-at CKGSB’s Marketing Research Forum.
Zhang Kaifu, Assistant Professor of Marketing at CKGSB, sat down with Little for a conversation on the evolution of marketing science in the US, and what Chinese companies can learn from it. Excerpts from the interview: Zhang Kaifu: When you started working on this subject, marketing was not considered a science. John Little: No, but everything has fundamentals. And they remain to be discovered and utilized (in practice). In the 1960s, I worked on media models, for example, and there is certain kind of data that exists and you can build models which will use (that data), but there are also judgments (that) have to be made. And one of the things people have been reluctant to do was to call it a science when you are using judgments.
ZK: So they consider science to be a hard, precise and mathematical process where there is no human judgment involved. JL: Well… that’s inappropriate in managerial conditions and in marketing conditions, usually. And therefore I wrote some papers which use judgments, and as time has gone by, we have understood processes better and better and require less (judgment) in some sense. I have a paper called “Managers and Models” (on how managers can use these models).
And for managers to be willing to use something, it has to be fairly simple, it has to include the important elements that he perceives as problems.
ZK: Not too complicated so they still can comprehend the essence… JL:That’s right. So a lot of people like that idea. Prior to that, you couldn’t do any marketing science or management science without a very, very complete database, or they had a database so they analyzed it but they left out important things.
ZK: That is the beginning of marketing science. And as we know that after 40 years, many firms, in the US at least, have already adopted many tools from marketing science-conjoint analysis, optimization, etc.-and many of them have already proved to be quite successful. We want to get a little of that historical perspective from you. How did this adoption process first start? JL: I founded a consulting company with some colleagues. The academics wanted to solve new problems and publish them, but they then hired MBAs (to put this research into practice).
We were willing to do the methodology and see that it worked. And one of the interesting things that I was involved in was building a marketing mix model at Nabisco which (was then known as National Biscuit Company and is now owned by Kraft) for Oreo cookies and another one for Coca Cola.
ZK: So it’s like a measurement/optimization tool? JL: That’s right. And perhaps the most well-modeled stuff-which I’m not so much involved in-is new products, because so many products fail. My colleague Glen Urban (David Austin Professor in Management, Emeritus, at MIT) also worked out a measurement process that you can take into a mall, and get people to try the product, and also to view potential advertising for the product.
ZK: This can be done before the product launches and it can help make some improvements… JL: That’s right. And it used to be. The product launch cost million dollars and this process cost maybe $60,000, so the payoff was huge. And if you are going to kill it, kill it early.
ZK: That’s all for the better. You mentioned that companies like Nabisco and Coca-Cola adopted some of the marketing science models early on. Did you see an impact after the companies made the adoption decision? JL: Well, interestingly enough, we had a built-in process of adaptation. It is (to see if) advertising effectiveness decreased or increased. Then you can shift the marketing mix and that sort of thing is very helpful. In fact, you can do experimentation.
ZK: Counterfactual simulations before you actually carry it out in the real world… JL: That’s correct. But you can also carry it out in the real world through a continuing re-measurement, because you introduce different advertising copy. The market changes, and you introduce different promotional techniques.
ZK: Did some of these insights translate into actual managerial actions? Did Coca-Cola, for instance, change its advertising strategy after that? JL: Yes. Actually what one discovers when one picks up and looks inside is that Coca-Cola does not sell soft drinks. They sell syrup.
ZK: And the meaning attached to it. JL: That’s right. So they are very seriously interested in advertising since their bottlers do most of the physical work and distribution.
ZK: And your company for the first time quantified the effectiveness of advertising? JL: That’s right. Anyway, it’s grown from there. And one of the things where I hit a jackpot early from a marketing science point of view was discovering with a student of mine, Peter Guadagni, that the logit model, multinomial logit model (Editor’s note: these are models that predict market share based on product and consumer characteristics), was remarkable. I’ve run millions of regressions and they always look bad, but these models look really nice. And this payoff would end in ways that can be used. For example, one of the promotional techniques frequently used in the US is to give coupons which you sometimes get in a store, sometimes get in the mail. And if you present them at the checkout, they entitle you some ten cents off or a dollar off, depending on the kind of product you buy. And for those we use the logit model to predict ahead, to give a baseline as if the coupon had not been dropped. Then it was dropped and you can see a jump in the baseline.
One of the things about knowledge is that your first few (findings) are your maximum discovery. After that, they all seem to be almost predictable, in the same ballpark (range). And so people lose their interest. Let’s say they have a fancy new idea, in which case you can have a new “aha!” experience.
ZK: But it is the major discoveries that you mentioned that are really pushing the boundaries of knowledge. JL: That’s right.
ZK: This is really one of the major issues that is attracting a lot of attention in China nowadays. Many managers are asking whether they can they really apply marketing science tools and make their business processes more efficient. In a way, China today is a little bit like the US 30 years ago, in terms of the application of marketing science tools. And I think the examples that you’ve just given us provide a confirmative answer. JL: And you have to have somebody do the work who has the skills to do it now. We found that many of our students-again we have mostly MIT people-but also a sizable number of them from Wharton, who understood what we were doing and could translate it. And you got to translate it.
ZK: In a way, MBA students are really like a bridge between academics and practitioners. JL: Exactly. And we turned out a lot. If you think about the process of building a model or doing a project like this, first of all, you have to have entry into the company. And you’d like to get a pretty high level because you frequently need a champion as it’s not going to be done immediately. It’s going to cost some money.
ZK: At least, it’s a one-of-a-kind problem, so you probably need a lot of communication between the company and the developers. JL: Absolutely. So you go in, and you go in with priors usually, a feeling, a hypothesis about what you might do. And then you do a lot to update that because you get the manager to articulate the problem. You go and talk to other people. Then you may come up with a model of some kind. But then you need data and you need to test the model and be sure you are doing it. And then one of my colleagues says that you should ask the manager, or discuss with the manager how he and you are going to evaluate the results, what is the success.
ZK: I presume that by success, you mean both for the company and in terms of knowledge creation. JL: That’s right. The company isn’t to worry about that, which is okay. And I will say that the companies have very complicated objectives sometimes. But I usually say that if you can increase their profits, they can figure out what to do with them.
But then you go on. And you have to… You learn by trying to fit the model with the data, but the model is wrong, and the data is probably right. Or you mean you may need different kind of data. It’s a circular process.
Zhang Kaifu is assistant professor of marketing at CKGSB. He holds a Ph.D in Management from INSEAD. His most recent research explores both theoretical and empirical issues related to media, advertising and the Internet.