“Marketing is still an art, and the marketing manager, as head chef, must creatively marshal all their marketing activities to advance the short and long-term interests of their firm.”

Neil Borden

Before diving into marketing measurements, it’s essential to understand how marketing models came to exist. As with any subject, we can trace the evolution of marketing measurement to understand better the trials and tribulations of those who came before us. With this, we can leverage what is still relevant, and discard anything which may not be relevant any longer. Understanding the history of the field also gives us a solid foundation upon which to build and, hopefully, a deeper and richer understanding of the subject.

The term marketing mix was coined in the early 1950s by Neil Borden, a Harvard Business School professor. Professor Borden gained inspiration for the term from a colleague [1] named James Culliton, who wrote in a paper that, “the business executive [is] a ‘decider,’ an ‘artist’—a ‘mixer of ingredients,’ who sometimes follows a recipe prepared by others, sometimes prepares his recipe as he goes along, sometimes adapts a recipe to the ingredients immediately available, and sometimes experiments with or invents ingredients no one else has tried.”[2]

Even though Borden coined the term in the 50s, he wrote a paper in 1942 entitled The Economic Effects of Advertising. Borden concluded that the success and necessity of advertising are directly related to the goods sold. For commodities, promotion has little effect in molding consumers’ general consumption patterns. But for a product like toothpaste, the cheap, unadvertised brand can make little headway because the consumer has a minimal idea of what they are buying, and the reputation of the advertised brands makes one fear that the cheap brands are of inferior quality or even damaging [3].

In a review of the paper written by the brilliant economist Joan Robinson, she states that, “All conclusions about the effects of advertisement must be tentative because of the impossibility of separating advertisement from the other influences upon demand and upon costs which are constantly changing along with it, but some broad generalizations are possible to make.” [4] Keep in mind that, at the time Robinson wrote this review, economics was still more theory than applied. (It wasn’t until 1969 that econometrics became a field, with Ragnar Frisch and Jan Tinberg defining the term and subsequently earning the Nobel Prize for their work.)

Other people from different disciplines were also tackling marketing and marketing measurement–namely, operations researchers. 

Quantitative marketing measurement, as we are familiar with it today, has its roots in operations research. For the uninitiated, operations research is a field that was developed out of necessity during World War 2 by the British military.

Operations research is concerned with the application of advanced analytical methods toward decision-making problems [5]. One of the first practical examples of operations research in practice was a group of scientists who figured out how to leverage a small amount of radar equipment for tracking the German Air Force.

Prior to the work of these scientists, placing radar sites involved a “guess and check” methodology, which is time-consuming and resource-intensive. This small team used math to find the optimal locations for their radar equipment, enabling efficient communication between the radar sites and military bases. The implication was faster repair times and more excellent sky coverage, leading to better axis movement tracking, fewer equipment losses, and decreased casualties. The United States military saw the efficiencies that the British intelligence gained by leveraging operations research and started implementing similar methodologies for various uses. [6]

Many of these talented researchers turned their interest toward other problems once the war ended. Luckily for marketers, some of these researchers focused on marketing.

One of those researchers, John F. Magee, published one of the first papers on marketing measurement in 1953 in the Journal of the Operations Research Society of America [7]. This paper, The Effect of Promotional Effort on Sales, is concerned with uncovering the relationship between sales volume and promotional effort. The research aimed to reveal how much promotional activity is justified to drive sales. It’s funny that 70 years later, we’re still discussing the same issue.

Using a beautifully simple model, Magee takes monthly sales data and can isolate the effects of promotional activity. Using this model, he can determine the amount of sales resulting from any given level of promotional effort. To contextualize this period, we can note the invention of the IBM 701, COBOL, and the creation of the first compiler. Suffice it to say; it’s a beautiful paper with applications almost a decade later.

Magee uses probability theory and some calculus in this paper to define three distributions that describe an average, higher-performing, and low-performing store. He then takes these formulations to isolate the impact of promotional activity. The paper is worth the read, but the main takeaway is that at this point, Magee was comfortable with a high level of randomness in his formulation. He aggregates by monthly data, doesn’t include seasonal activity or other covariates, and ignores times altogether. The key definer of this generation of work is the use of simpler analytically solvable probability models over more complicated alternatives. Sometimes, simple models work wonderfully. 

Four years later, researchers published a new paper examining advertising from a more holistic perspective. The management consulting firm, Arthur D. Little, had been doing a lot of work trying to understand how they could evaluate the effectiveness of an advertising campaign and present some of their generalized findings in the paper, An Operations Research Study of Sales Response to Advertising [8].

Arthur D. Little had a lot of clients and the ability to conduct numerous tests all over the United States; because of this, they were able to create a model that fit their experimental data. The authors compose the model utilizing three main concepts: the sales decay constant, saturation level, and response constant. The sales decay constant captures their observation that sales decrease when marketing activity is limited. This decrease is directly related to numerous issues, including having an old product, other competitors, and other factors. The concept of saturation level implies that consumers of the product become saturated after sufficient spending, and increasing expenditure results in diminishing marginal returns. Lastly, the response constant describes sales when the product is initially measured.

Similar to the paper by Magee, this paper aggregates sales into years, so it gets to ignore some level of seasonality. Still, it does “link” years together through the concept of sales decay. This paper is also the first paper (that I could find) that introduces the idea of saturation effects, which are now commonplace among measurement methods. 

In 1969, Jean-Jacques Lambin wrote the first regression-based paper I found, Measuring the Profitability of Advertising: An Empirical Study [9]. This paper comes from an econometrics frame of mind compared to the two prior papers written by operations researchers. As such, some of the nomenclature changes. Lambin formulates a demand function that uses sales, disposable income, weather, goodwill, advertising spending, visit frequency change, and retail price change. Most of these things are intuitive to include, except “goodwill,” which is the delayed impact of marketing.

Lambin leverages the work from The Measurement of Cumulative Advertising Effects, written by Kristian Palda. Palda’s work parameterized the cumulative effect of advertising over time utilizing distributed lags—the first paper I found to do so. The paper from Lambin is the first representative I found that utilized regression analysis. They used an IBM System/360, a computer released in 1964 weighing between 1,700–2,310 lbs. Additionally, it’s important to note how this is one of the first papers where econometricians start to add their “flavor” to media measurement. It’s no coincidence it was published in the same year the field of econometrics was founded.

At this point in history, we have all the building blocks of a modern media mix model. The techniques I’ve mentioned are not without their tradeoffs, and in recent years, we have improved how we can put them together. Other papers certainly expand on these methods and add their permutations (finding these can be a fun exercise for the reader). Luckily, we can now stand on the shoulders of giants and leverage modern computation to improve the accuracy and reliability of our models.

Key Points:

1. Sales go up and down depending on spend. Sales might fluctuate at different rates (it’s not linear).
2. How sales relate to advertising can take any number of shapes, including the common sigmoid S-shape or even a concave shape.
3. The way your competitors advertise directly influences the health of your business.
4. Spending today can impact sales tomorrow.
5. How advertisements perform can vary due to many factors, like changes in the media, copy changes, or other factors.
6. Even if ad spending is constant, sales can continue to fall off. 

References

1. Borden, N. H. (1964). The concept of the marketing mix. Journal of advertising research, 4(2), 2-7.
2. Magee, John F. “The effect of promotional effort on sales.” Journal of the Operations Research Society of America 1.2 (1953): 64-74.
3. Culliton, J.W. (1948) The Management of Marketing Costs. Harvard University, Division of Research, Graduate School of Business Administration, Boston, MA 
4. The Economic Effects of Advertising. By Neil H. Borden. Chicago: Richard D. Irwin, Inc. 1942.
5. INFORMS. (n.d.). What is O.R.? Retrieved from https://www.informs.org/Explore/What-is-O.R.-Analytics/What-is-O.R.
6. Operations Research in World War II. https://www.usni.org/magazines/proceedings/1968/may/operations-research-world-war-ii
7. Vidale, Marcelle L., and HB88402 Wolfe. “An operations-research study of sales response to advertising.” Operations research 5.3 (1957): 370-381.
8. Lambin, J. J. (1969). Measuring the profitability of advertising: An empirical study. The Journal of Industrial Economics, 86-103.
9. Palda, K. S. (1965). The measurement of cumulative advertising effects. The Journal of Business, 38(2), 162-179.