By Kevin Gray and Michel Wedel
Kevin Gray: There has been a lot of buzz about data science, big data, analytics and so on in the past few years, and a lot of marketing researchers seem confused about what all this means. Could you give us a simple layperson’s definition of marketing analytics?
Michel Wedel: Marketing analytics involves the examination of data about how customers feel, act, and interact around products and services, using descriptive, diagnostic, and predictive metrics and mathematical methods. Its purpose is to obtain insights into customer behavior and to improve the effectiveness of marketing performance. Marketing analytics is increasingly interdisciplinary, combining methods from business, mathematics, statistics, econometrics, psychometrics, and computer science.
KG: How did all this evolve? Could you give us a brief history of marketing analytics?
MW: The beginning of the systematic use of data in marketing is widely credited to Charles Coolidge Parlin around 1910, with his work on advertising for the Curtis Publishing Company in Boston. In the 1920’s and 1930’s the first marketing research companies, such as Nielsen, Burke and Gfk, were established. They have long used analytics to support their clients’ marketing decisions. One could argue that the use of big data analytics in marketing begun in the 1970’s, with the introduction of IBM’s point of-sale scanning devices for Universal Product Codes. This marked the onset of large scale capture and analysis of digital transaction data by marketers. The availability of digital data exploded when in in the 1990’s the World Wide Web came into existence (clickstream data), and Google was founded (search data). Then, from 2000-2010 things quickly evolved with the launch of Facebook (social network data), YouTube (video data), and the iPhone (location data). These new sources of data are now widely used for analytics by marketers, which along with the analytical techniques developed for these data, has given rise to many entirely new forms of marketing.
KG: Flashing back a decade – to 2007 – what was the buzz then? What were experts predicting about marketing analytics then that they basically got right, and where did they go wrong?
MW: Although the term big data had been used before, around that time it began appearing regularly in scientific articles, blogs, Google search terms, and job postings. It is now clear that initially, many businesses’ expectations about the potential of big data were overhyped. Much of the initial surge in technology evolved around data storage, and companies invested too much in data capture, without concise plans for how to the data should be used to improve marketing decision making. Moreover, investments in analytics capabilities lagged behind. Today, the success of industry leaders such as Amazon, Google, and Facebook has made it clear that the potential of big data for marketing decision making can be leveraged only through the use of analytical tools. In addition, it is becoming clear that the availability of big data itself may give rise to data-driven decision cultures in companies. Analytical tools have been shown to provide companies with competitive advantages, and to positively impact their financial performance.
KG: Thinking about data and analytics now, in 2017, what should marketing researchers and data scientists focus on most?
MW: Historically, the development of marketing analytics has progressed through three stages: (1) the description of observable market conditions through descriptive statistics and dashboards, (2) the development of statistical and econometric models as diagnostic tools, and (3) the evaluation, optimization and automation of marketing decisions. New, unstructured digital data in the form of blogs, search results, reviews, images, locations, video and tweets enable deep and actionable insights into the economics and psychology of consumer behavior. But, the usage of these new data sources in marketing practice is still mostly in the first stage of its development (description), and its full potential remains to be tapped. Machine learning methods such as deep neural networks and cognitive systems have ushered in the second stage of analytics for big data, and are becoming more and more popular in practice. The challenge is to combine and utilize all these sources of unstructured data to optimize and automate marketing decisions. Developments in practice have mostly involved “small stats on big data”, while academic research has used “big stats on small data”. Collaboration between researchers in academia and in practice is needed to develop methods that solve important marketing problems with an eye for application and computation.
Key areas of development involve: (1) how to include new rich data in marketing mix models to improve explanations and predictions of the marketing effects; (2) how to attribute elements of the marketing-mix to various touchpoints in the customer purchase funnel; (3) how to dynamically allocate marketing resources across various offline and online channels and multiple devices; (4) how to assess causal effects of marketing control variables through analytical methods and field experiments; (5) how to personalize marketing mix elements in fully automated closed loop cycles; and finally (6) how to best apply analytical methods to protect data security and consumer privacy.
KG: And, thinking a decade or so ahead, can you offer any thoughts on what marketing analytics might look like then? Do you think there are major surprises in store for marketing researchers and data scientists?
MW: In my view, two of the most important developments in the coming decade are the Internet of Things (IoT) and Natural User Interfaces (NUI). Well over 15 billion devices already have sensors that enable them to connect with other devices and transfer data without human interaction. Add to this the trend towards wearable devices which increasingly collect physiological data and communicate with other devices; in some cases blurring the line between the consumer and the devices. The IoT has already started to bring the offline world online, and thus open up many offline behaviors and interactions to the same type of analysis as are now possible for online behaviors. The IoT will change the way people interact with their man-made environment, become a major source of new product and service development, and generate massive data in the process. In addition, the rapid development towards NUI enables people to interact with their devices through voice, motion control, gaze, facial expressions, and even in some cases through mere thought. Thus, people are increasingly interacting with devices as if they were other people. This will change the nature of the interactions, and as data on speech, eye movements, facial expression and motions are recorded and available to marketers at massive scales, this will necessitate the development of tools for their real-time analysis. This will open up entire new ways for product and service customization and marketing-mix personalization, but very little work has addressed the analytical requirements for these developments as yet.
KG: Lastly, when looking to hire analytics staff or subcontract analytics work, what are the most important considerations? What should clients be looking for in an analyst or analytics company?
MW: Marketing analysts typically work at the interface of statistics, computer science, and marketing, and they need to have broad and deep skills. In addition, areas of marketing such as advertising, product development, search marketing, segmentation, each have their own requirements for data and analytics. Analysts therefore need deep knowledge of marketing modeling, marketing-mix optimization, and personalization, and must be able to apply state-of-the-art statistical, operations research and machine learning methods. People with only excellent technical skills are often not as effective in companies as people that also have significant domain knowledge, because they know how to interpret data and findings in the light of extant knowledge on marketing and consumer behavior. Moreover, successful analysts are capable of effectively communicating the results obtained from analytical techniques to decision makers.
Business leaders believe that the difficulty of finding talent with these skills is the main barrier toward implementing big data analytics. In successful companies analysts are often cultivated through continuous on-the-job training. The education of marketing analysts with such a broad and deep skill set has posed a challenge for business schools. Next to existing specializations in undergraduate and MBA programs at many universities, recently created masters programs in marketing and business analytics focus on developing these multidisciplinary skills in students who already have a rigorous training in the basic disciplines. These programs offer great promise. Closer collaboration between universities and companies is needed, however, to make sure that educational programs remain relevant to the requirements of the industry.
KG: Thank you, Michel!
For a detailed look at marketing analytics see Wedel, M. and P.K. Kannan, 2016. Marketing Analytics for Data-Rich Environments. Journal of Marketing, 80 (6), 97-121.
Kevin Gray is president of Cannon Gray, a marketing science and analytics consultancy.
Michel Wedel is Distinguished University Professor, University of Maryland. He has received the Dr. Hendrik Muller Prize for outstanding contributions to the social sciences from the Royal Dutch Academy of the Sciences, the Charles C. Parlin award for exceptional contributions to Marketing Research from the American Marketing Association and the Churchill Award for lifetime contributions to the study of marketing research from the American Marketing Association. He is a fellow of the American Statistical Association and the Institute for Operations Research and Management Science.