Where is Marketing Data Science Headed?
Marketing data science - data science related to marketing - is now a significant part of marketing. Some of it directly competes with traditional marketing research and many marketing researchers may wonder what the future holds in store for it.
Many associate “data science” with self-driving cars, artificial intelligence, smart cities and bioinformatics, though it is increasingly part of marketing. Marketing data science is multifaceted but at its core is personalized marketing, a high-tech extension of direct marketing. Two manifestations of it, personalized ads and recommendations, are now so ubiquitous that they often escape our notice.
Introduction to Algorithmic Marketing (Katsov) and Digital Marketing Analytics (Hartman) are two books that provide excellent overviews of how data science is used in marketing. The Data and Marketing Association website may also be of interest to many readers, and the JM-MSI Special Issue: "From Marketing Priorities to Research Agendas" offers a deep dive from an academic perspective.
As is anything, marketing data science is affected by exogenous factors such as economic downturns and technological innovation, not to mention pandemics. Privacy legislation as well as the popularity of ad blockers and search engines which do not track are having an impact. A "cookieless future" is already the subject of blogs and articles.
Another topic that can become quite heated is the efficacy of personalized marketing in the first place. Some maintain it doesn't really work from a technical standpoint, that is, that personalized ads and recommendations do not actually reach the right people at the right time, or the right people at any time.
Part of this is a data issue. Not a small amount of the “big data” used in digital marketing consists of imputed and aggregate data that have been mashed together. Marketing scientists and others who analyze and model data know that missing data and even small errors in a single data file can wreak havoc on our analytics.
Of course, this is just one point of view and many will contend that personalized marketing is more effective than mass marketing or the macro segmentation most familiar to marketing researchers.
Predictive analytics as currently practiced is mainly concerned with the What (which includes where, how, by whom, and how often) but not the Why. Understanding why consumers behave as they do can help marketers predict their future behavior, communicate effectively with them, and design products they'll like. Inferring why they behave as they do from what they've done in the past is not always straightforward because different people may do the same things for different reasons. They might also do different things for the same reasons.
A simple example would be household cleaners. One person might buy Lysol Clean most often because of its disinfectant properties, and another because it’s easy to find where they usually shop. One person might use Goo Gone because they like its fragrance, and someone else Citrasolv because they like its fragrance.
Another concept sometimes called the multiple me also comes into play. This notion reflects the simple fact that consumer behavior often varies according to occasion or motivation - buying wine for cooking versus buying wine to celebrate a promotion being one illustration. Our behavior can also vary with time due to changes in personal circumstances, such as marriage, birth of a child, a promotion or relocation. Our tastes may change too.
As a result, when the Why is overlooked much is missed.
Other critics say personalized marketing frequently subsidizes purchases that would have been made anyway and, in the process, increases price consciousness among consumers and, over time, can erode brand equity. Smart marketers, indeed, would be wise to avoid encouraging deal shopping. Les Binet and Peter Field have much to say worth listening to regarding these topics.
Some maintain that personalized marketing usually works but isn't always cost effective. Marketing ROI can be tricky to estimate since some costs are guestimates, and adjusting for exogenous factors such as recessions and competitor activity is challenging - competitors’ marketing, for example, can actually benefit our brand if it helps grow the category.
Moreover, it may appear cost-effective if the investment in data infrastructure and analytics teams are ignored. Cost-effectiveness of targeting can also be deceptive when we have a higher “hit” rate among a much smaller number of consumers than without targeting. Overall sales may actually decline with targeting even when marketing ROI is higher.
Still others, notably Byron Sharp and his colleagues at the Ehrenberg-Bass Institute, feel targeting as a rule is a bad habit and a bad idea. How Brands Grow (Sharp), How Brands Grow: Part 2 (Romaniuk and Sharp), and Building Distinctive Brand Assets (Romaniuk) in my opinion are must-reads for marketers and marketing researchers.
There is also the issue of ad fraud, which is not trivial and now harder and harder to sweep under the rug. Bob Hoffman and Augustine Fou are good sources for more information on this touchy subject.
I have not mentioned UX (User Experience) or CX (Customer Experience), which are related to marketing and have benefited from having more data, richer data and more sophisticated analytics tools. I should also point out that predictive analytics and other aspects of personalized marketing can be enhanced by traditional marketing research methods such as focus groups and consumer surveys. It’s not either/or.
These “old” methods, which continue to evolve, can be immensely helpful in shedding light on the Why. Awareness of this synergy is growing among marketing researchers, if not among data scientists, who are often focused on data management and programming. I’ve long felt this was an overlooked opportunity for marketing research.
Although I have mostly emphasized threats to marketing data science in this article, as a marketing data science person myself my own biases mostly run in the opposite direction. However, as a businessperson I cannot afford to believe in panaceas or just believe what I wish to believe. Even before the pandemic, I sensed growing skepticism in the business community and among investors regarding AI and the value of big data - see Big Data, Big Dupe (Few) and laughing@big data (Jones) for examples of what I mean.
So, to sum up, I don't have a clear answer to the question "Where is marketing data science headed?" and I'm not sure anyone has. I just try to be careful about making assumptions and taking too much for granted.
Bio: Kevin Gray is President of Cannon Gray, a marketing science and analytics consultancy. He has more than 30 years’ experience in marketing research and data science with Nielsen, Kantar, McCann and TIAA-CREF.
Original. Reposted with permission.
Related: