Increasing Customisation: A Revealed Preference For Data Science

As larger swathes of the global population have come online each year, products and services everywhere in the world have followed them there. In India, growing at breakneck pace, the e-commerce industry will be worth over $100 billion by 2020, driven by an expected surge in the number of Internet users to 720 million, which is itself growing at over four times the global compounded annual rate. An important implication of millions of goods and services being bought and sold online is that all of these transactions are recorded systematically, and lend themselves easily to statistical analysis.

Complemented by ever-growing computing power and increasingly sophisticated data mining tools, this has significantly enhanced our ability to better understand consumer behavior, predict it more accurately and service it with greater nuance. This has, in turn, fuelled enhanced efficiency and greater customization in every industry, from fashion to education.

For example, we recently chanced upon a website e-tailing formal office-wear shirts, formerly the most “vanilla” items of clothing in men’s wardrobes. It offered scores of fabric options, about a dozen collar and cuff designs (including one worn by Sean Connery’s James Bond, no less!), another twenty-odd button choices and even the option to have one’s initials embroidered onto the cuff, all while simulating digitally how each choice would affect the way the shirt looked. Imagine how much this company would know about our shirt preferences if we were to order something from them! Using data science and predictive modeling, they could then custom-design shirts based on our “revealed preferences” (Samuelson, 1938), and have them follow us everywhere we go online through search engine remarketing. Ceteris paribus, or other things remaining the same, that would make it much likelier that our subsequent shirt purchases too are from the same website, wouldn't it?

A 2015 McKinsey report found that personalization, or customization, can increase sales by at least 10%. Besides, on account of being low-cost and following a test-and-learn self-improvement template, it can deliver five to eight times the usual return on marketing spend. Therefore, as consumers have evolved to expect greater and greater customization, firms have re-aligned their marketing systems and strategies to stay ahead of the customization curve by employing sophisticated tools, techniques, and personnel for analyzing the copious amounts of data collected. This has placed a huge premium on the skills of data scientists and business analysts, who apply their knowledge of statistics, computer science, economics, mathematics and business to dissect this Big Data for corporations, and discern consumers’ expected willingness to avail of an offer, use a new product feature, consume some content or buy a product.

The supply of people with these skills is still limited in the market, with demand far outstripping supply, and fittingly this presents a customization opportunity of its own for higher education institutions. The need is for institutions to customize the traditional silos of the aforementioned subjects into an inter-disciplinary product or programme which trains and prepares skilled data scientists and business analysts. These graduates would be skilled at handling large datasets and the complex mathematical and statistical models used to analyze them while being able to draw valuable commercial insights allowing them to hit the ground running and add value in the world of work. 

The market value of graduates with such a skill set is not limited to marketing and the corporate sector. In fact, the application of data science holds enormous promise for policymaking. In the short run, data scientists can help governments understand the impact of nationwide social sector schemes and improve the targeting of subsidies. In the long run, on the other hand, the implications could be as profound as re-shaping how economists measure utility or welfare, both for consumers and for society as a whole, basing it not on what consumers say, but on their preferences as “revealed” by what they do, that is to say, by the “My Orders” sections of all the e-commerce websites they use. 

In fact, as the world becomes ever more data-driven, career options even for data scientists are only limited by one's imagination. The rewards for data science graduates are clear from market and business trends. In addition, the current, and widening, skills deficit acts as another compelling market signal that we need workers who can blend their knowledge of mathematics, statistics, computer science, and management, to mine huge datasets and filter out the signal from the noise. Growing customization provides one such obvious and lucrative career path, whereas we use data science to better target products and services based on consumers' preferences, a firm's bottom line will inevitably improve! 

Several higher education institutions teach one or more of the disciplines underpinning data science anyway. The question is whether they can adapt to current market needs, by themselves offering a customized, inter-disciplinary education product.

James Abdey

Guest Author Professor James Abdey teaches the Department’s large service-level undergraduate courses in mathematical statistics and quantitative methods, as well as elective courses in market research. His research interests include market research techniques and forensic statistics – the interplay of Statistics and Law. Prof. Abdey (Assistant Professorial Lecturer, LSE & Guest Faculty and Faculty Trainer at ISBF) has been closely involved with LSE’s Summer School and the University of London International Programmes for a number of years. His association with ISBF has grown over the last few years and he’s travelled to India to conduct Faculty Training and Teachers’ Symposium over the past two years.

Chiraag Mehta

Guest Author Chiraag Mehta is an Associate Professor at ISBF, New Delhi