Andrew Ehrenberg (1926-2010)
Andrew Ehrenberg was an incisive statistician who applied statistics to the study of marketing phenomena. More specifically, to understand consumer buyer behavior and how advertising works. Andrew’s contributions to modeling and methodology in marketing are seminal.
Andew attended Queen’s College, Taunton, and then studied statistics at Kings College, Newcastle-upon-Tyne and Cambridge. In 1951 he became a Lecturer in Statistics at the Institute of Psychiatry in London, and in 1955 moved into commercial marketing research and consulting. From 1970 to 1993, Andrew served as Professor of Marketing at London Business School (LBS), where he also set up and ran the industry-funded Centre for Marketing and Communication (CMaC). In 1993 Andrew became Professor of Marketing at the London South Bank University.
The Ehrenberg-Bass Institute summarizes Andrew’s contributions thus: “Ehrenberg’s fundamental belief was quite simply that the methods of physical science are also applicable to the social sciences. This principle enabled him to establish wide-ranging empirical quantitative generalisations about human behaviour.”
Andrew’s most celebrated and perhaps most impactful contribution is the application of the Negative Binomial Distribution (NBD) to consumer behavior. In demonstrating the compelling applications of the NBD, and in enriching the model structure over the years, Andrew had a purposeful intellectual partner in Gerald Goodhardt. Andrew and Gerald were prolific co-authors. Their partnership lasted for almost 50 years beginning in early 1960s through 2010. In fact, one of the first and seminal publications showing the application of the NBD to consumer buyer behavior is their joint publication in Journal of Marketing Research in 1967 (Conditional Trend Analysis: A Breakdown by Initial Purchasing Level, JMR, Vol, IV, 155-161.) Through their many empirical studies, Andrew and Gerald found this robust empirical generalization: the distribution of purchase frequencies of a brand over any fixed timescale follows a predictable pattern that can be accurately modeled using the (rather flexible) NBD.
Andrew’s impact on model structures in marketing have been deep. For his sustained and impactful contributions, Andrew was awarded the Gold Medal of the British Market Research Society twice, first in 1969 and again in 1996. He was also recognized with the Honorary Fellowship of the Royal Statistical Society. In December 2005 he was awarded an Honorary Doctorate by the University of South Australia.
There are many scholars whose work reflects Andrew’s thinking. But none captures the breadth and depth of Andrew’s work as does Donald (Don) G. Morrison, Professor Emeritus at University of California at Los Angeles, an ISMS Fellow, a contemporary of Andrew, and a friend (sometimes argumentative) of Andrew. Not only in Don’s work but also in many of his celebrated doctoral students, Andrew’s impact is evident.
So, let us listen to what Don says. But for Andrew and his colleagues, Don would not have done that much research in stochastic models. “During the 1992 Marketing Science Conference in London, I got to meet and know Andrew's co-author Gerald Goodhardt. We swapped stories about working with Andrew. I probably would not have done as much in the stochastics models area if not for Andrew and his colleagues.”
Don delights in describing Andrew’s insights in stochastic models, and how Don got an opportunity to expand on those insights. “Andrew was the first to use the Negative Binomial Distribution in marketing, but it was used decades earlier in studies of accidents. I generalized the NBD by including a group of never buyers. I also relaxed the exponential distribution of interpurchase times. One of my papers that Andrew loved showed that if you include hard core non-buyers and also make the interpurchase times more regular than exponential, these two effects almost cancel each other and make simple NBD work.”
There is more from Andrew: contributions to methodology and analyses. Here is Don describing so eloquently Andrew’s insights into data analyses. “Andrew wrote some classic articles on data analysis. One example I used for decades in my multivariate statistics course involved a 10x10 correlation matrix. On the first page a principal components analysis was done. The 10x10 matrix had the correlations. It was a mess. The next page reduced the correlation to one significant figure and rearranged the rows and columns. Now looking at this there were two clear segments, and they were virtually orthogonal I would then tell the students: "Just because you can do complicated analyses, look at the date first and try some simple analyses.”
Patrick (Paddy) Barwise, Emeritus Professor of Management and Marketing at London Business School, co-author with Andrew and a friend and colleague of Andrew, offers this valuable insight on how Andrew conceptualized and approached research. “Andrew thought all academic marketing research should start with empirical generalizations (EGs) and only then look for theory. His own theorizing was about stochastic models that fitted the data, notably the Dirichlet model of multi-brand repeat-buying (Goodhardt, G. J., A. S. C. Ehrenberg and C. Chatfield, (1984), ‘The Dirichlet: A Comprehensive Model of Buying Behaviour’, Journal of the Royal Statistical Society, Series A 147, 5 (1984), 621-655.) The Dirichlet parsimoniously accounts for a vast number of observed variables with remarkable precision. But, by excluding causality, treating market share as an independent variable, and assuming a steady-state market, it doesn’t directly tell marketers what to do. Instead, Andrew’s main message to them was that brand loyalty was weaker than they thought; that, over (almost) any time period, the main difference between big and small brands was the number of people or households buying at least once, not their average purchase frequency; and, therefore, that they couldn’t build a ‘niche’ brand based on a few ultra-loyal buyers. This is an example of McPhee’s Law of Double Jeopardy’ (McPhee, W., (1963), Formal Theories of Mass Behavior, Free Press). Less popular items are not only chosen or mentioned by fewer people than more popular rivals, but also chosen slightly less often and, on average, liked slightly less, even by those who choose or mention them at all.”
Mark Uncles, Senior Deputy Dean (Education & Student Experience) and a Professor of Marketing, University of New South Wales Business School, and long-term colleague of Andrew affirms and expands on Paddy’s observations on Andrew’s philosophy of science: “Andrew was a true scientist. Whatever the subject, his preference was to start with observations and data. He looked for patterns, trends and regularities, from which he might derive empirical generalizations. Any exceptions or anomalies would be noted. This provided the basis for descriptive understanding of phenomena. Typically, the understanding would be summarized in simple and elegant mathematical expressions and models. Equipped with empirically grounded models, managers had the tools to make predictions and inform their decisions. In summarizing what was currently known, Andrew would also highlight what was not (yet) known. Scope, and boundary conditions, were as key to his approach as the identification of patterns, trends and regularities. He was deeply sceptical of approaches that started with limited data and elaborate theorizing. He was scathing about over-complex models. Why would you want elaborate theory or complex models when you weren’t even sure there was a phenomenon to describe, let alone explain? Why concern yourself with outliers before being clear about the generalizability of the main patterns? How can you jump to a decision-making model without first having a well-informed descriptive understanding of the phenomenon of interest? At the time these views seemed to be bucking the trend because academic papers were becoming more technically sophisticated.”
Mark Uncles summarizes Andrew’s intellectual contributions eloquently thus: “Andrew’s academic interests and accomplishments ranged across numerous fields of study. As an insider, I was very aware of this. I’m not sure this was ever fully appreciated by the broader academic community during his lifetime. Significant intellectual contributions were made in areas as diverse as buyer behavior, store choice and market structure, audience research and television viewing, advertising and pricing research, statistics and data analysis, marketing management and public policy. He could move with ease from describing consumer behavior in-store to viewer behavior at home, from analyzing behavioral data to interpreting attitudinal data, from simple stats to comprehensive mathematical models. The range of interests is remarkable enough, but even more because of the level at which he operated. With colleagues, he published in all the major marketing and statistical journals: Journal of Marketing, Journal of Marketing Research, Marketing Science, Journal of the Royal Statistical Society, Biometrika, The American Statistician, amongst many others. Also, Nature. And books: notably Repeat-Buying and Data Reduction. And found time for book chapters, reports, op eds, commentaries and letters to the editor. Andrew understood that whilst reputation might be solidified by publishing in top-tier journals, impact means harnessing multiple channels to reach influential audiences.”
Why was Andrew such a remarkably impactful global scholar in marketing? For two reasons.
One reason was that Andrew did not flinch from arguments, but he also was gracious to build friendships as argumentative as he may have been. Listen to Don on this, “Andrew Ehrenberg and I had many years of friendly battles. As you can imagine Andrew and I were often referees on each other's manuscripts. While officially blind, we of course always knew the identity of the reviewer. We were never enemies, and in fact were close friends. In fact, at conferences where we both presented, we had a kind of locker room banter.”
Another reason was his open-mindedness, and his staggeringly stubborn belief that science and discovery, and good intellectual output did not recognize barriers of age or gender or formal education credentials. Mark Uncles speaks to this eloquently in a personal and poignant manner. “We first met in 1985. I was completing my PhD thesis at the University of Bristol in a Geographical Sciences Department and the opportunity to work with Andrew was brought to my attention. I travelled to London to meet him at London Business School. We talked. I met close collaborators Gerald Goodhardt and Paddy Barwise. I would go on to join the Centre for Marketing & Communication (CMaC), based at London Business School, and spend almost a decade collaborating with Andrew and his colleagues.
The way we started to work together says much about Andrew’s outlook. Age didn’t matter. I was twenty-four, Andrew several decades my senior, and in the next few years those who joined CMaC ranged from young doctoral students to retired practitioners working as research fellows – John Bound, John Scriven, Len England, Neil Barnard and many others. Such was our conviction that age shouldn’t matter, Andrew and I published a paper on this theme in the Journal of Advertising Research. Nor did disciplinary or professional background matter. There were market researchers, statisticians, maths grads, and me with a geographical sciences background. Many members of the team had worked in marketing as product managers or market researchers, but it wasn’t a requirement. Leading statistician, John Nelder, held a position in CMaC – at the time he was President of the Royal Statistical Society. Andrew had a remarkable capacity to draw into his circle a highly eclectic group of co-workers.”
Paddy affirms Mark’s testimony and illustrates how such openness on the part of Andrew has led to a productive approach to science and discovery. Paddy was part of that journey, and he was drawn into Andrew’s thinking as a doctoral student of Andrew. Paddy is sympathetic to Andrew’s philosophy and is mildly critical of over-indulgence of Karl Popper’s approach to science. Here is Paddy on this. “When I joined LBS, with a science and engineering background and an MBA, but no PhD. I initially found nothing especially odd about Andrew’s research methods. It was only later that I learned that no one else then looked for EGs in marketing. Instead, the rule was that you had to start with prior theory; develop new (qualitative) hypotheses; test these against some kind of (again, qualitative) null hypothesis; and then treat the findings as if a 95% confidence level meant there was a 95% chance that a strict replication study would get the same result. All this seemed even odder when I learned that, in practice, marketing studies were hardly ever replicated and that, when they were, the second study often failed to reproduce the first one’s results – certainly far more often than implied by its confidence levels (Evanschitzky, H., C. Baumgarth, R. Hubbard and S. Armstrong, (2008), ‘Replication Research in Marketing Revisited: A Note on a Disturbing Trend’ (June 29)). This, in my view, reflects an over-rigid interpretation of Karl Popper’s Conjectures and Refutations (Popper, K., (2002), Conjectures and Refutations: The Growth of Scientific Knowledge, Routledge Classics (first published 1963). Popper’s term ‘conjectures’ seems to me much looser than ‘hypotheses’.) The natural sciences are based on well-established – because replicated – empirical findings and new theory is usually a response to anomalies in these: science history from Aristotle and Ptolemy to dark matter and neutrinos is full of examples (An extreme ‘data-first’ example is Darwin, who started with years of careful observation and only then developed his theories - slowly and reluctantly because of the scary religious implications.)”
Gurumuthy Kalyanaram (G.K.), a faculty colleague of Frank Bass and a doctoral alumnus of MIT, and now an MIT Education Counselor, recalls this. “While Frank Bass may not have completely agreed with Andrew Ehrenberg on all aspects of his philosophy and insight, he was a great admirer of Andrew. Bass’ research output was substantially influenced by stochastic model insights of Andrew and more specifically, applications of Negative Binomial Distribution to consumer behavior. For instance, one of Bass’s seminal papers, The Theory of Stochastic Preference and Brand Switching (JMR 1974), employs the NBD and the related insights in its analysis. Bass’ research in the 1970s and 1980s carried the distinct imprint of Andrew’s insights and thinking. Andrew not only influenced Bass but also several of his doctoral students including Abel Jeuland. This portfolio of research included, Equilibrium stochastic choice and market penetration theories: derivations and comparisons (with Abel Jeuland et.al., Management Science 1976), A multiband stochastic model compounding heterogeneous Erlang timing and multinomial choice processes (with Abel Jeuland and Gordon Wright, Operations Research 1980), and An investigation into the order of the brand choice process (with Manu Kalwani and Moshe Givon, Marketing Science 1984).
Frank Bass also saw great value in Empirical Generalizations. In fact, Frank Bass and Jerry Wind organized a conference on Empirical Generalizations in Marketing in Wharton in 1995. They also published a Special Issue of Marketing Science (Volume 14, No.3) on Empirical Generalizations. I got to publish two manuscripts focused on Empirical Generalizations in Reference Price (with Russ Winer) and Order of Market Entry (with Glen Urban and Bill Robinson). In the introduction to the Special Issue, Bass and Wind wrote, “Science is a process in which data and theory interact leading to generalized explanations of disparate types of phenomena. Thus phenomena (empirical generalizations) are the building blocks of science.” Bass and Wind went on to write that identification of such generalizations was not only important for scholarship but also for practice. There was so much synergy between Andrew and Bass. So, it is so very apposite to have an Ehrenberg-Bass Institute for Marketing Science.”
Andrew’s impact on practice was substantial. He founded an industry-sponsored organization to promote academe-industry partnership. Mark explains this: “The public face of these industry relationships was CMaC, supported and funded by a consortium of industry sponsors. At any one time there would be several dozen supporters each contributing moderate funds and benefits in-kind: packaged goods multinationals, advertising agencies, gasoline corporations, television networks and publishers. The breadth of coverage was impressive, although tough to sustain. Added to which, a significant proportion of supporters were located outside of the UK. For Andrew, flying visits to New York and Cincinnati were common. Andrew also led a commercial consulting service, ASKE Research, operated jointly with Gerald Goodhardt and Martin Collins. Dozens of reports were prepared, typically with clients sharing data and ASKE researchers giving meaning to the data. This was data analytics in action – before anyone had coined the term.”
Mark captures Andrew’s personality so eloquently here. “Certain words readily come to mind when thinking about Andrew’s approach to work. Driven. Passionate. A polymath. Compulsive. He was ‘always on’ and didn’t hesitate to call colleagues any time of the day. Tenacious. A ‘revise and resubmit’ was always regarded as an acceptance. An outright ‘reject’ was something to be challenged. To outsiders he would come across as confident, some might say super confident. Certain about his facts. His clipped style of writing and presenting reinforced this sense of certainty. And for those who didn’t share the same worldview, Andrew was invariably seen as contrary and provocative. He was fully capable of talking to a room full of academics and pointing out the error of their ways. Nor did he shy away from talking to the media and expressing strong views on marketing, broadcasting and policy issues.
Behind the scenes things were different. He would heavily rehearse presentations. Papers were continually edited and revised, both for content and style. Andrew would agonise over single words and phrases. He might have seemed certain, but we would debate and discuss ad infinitum. Frequently he would stand at the office door for a couple of hours debating a point that was on his mind. Colleagues were used as sounding boards.
He was always interesting and interested, and entertaining. A master of one-liners, memorable phrases, and mnemonics.”
Andrew will be remembered for his immense contributions, including his very robust empirical and philosophical insights, his generous mentoring of scholars, and his impact on practice.
Don recalls Andrew’s purposeful argumentation and “friendly battles” to advance thinking. Paddy speaks to the learning. “I did my PhD (‘Mass Attitudes and Routine Choice Behaviour’) with Andrew part-time over eight years, drawing on our research on brand choice and TV viewing. I learned a huge amount from my twelve years working with Andrew - although it did mean that my early research training was rather idiosyncratic. Decades later, his thinking is still influencing my work.” Mark speaks to the generosity of the man. “He was a generous and open-minded mentor. Far more so than outsiders might have expected. I was privileged to be associated with Andrew.”
Gurumurthy Kalyanaram (G.K.)
Dated: December 2021