Tapan Khopkar, consultant with Cartesian Consulting Pvt. Ltd, has extensive experience and expertise in all phases on the analytics cycle. He has worked with Indian and international clients across verticals such as retail, e-commerce, travel, financial services, industrial products. Prior to joining Cartesian, he completed a doctorate in Information from the University of Michigan, Ann Arbor.

In this article, he tells you how you can answer the three big questions in data-driven marketing.

Marketing in the Age of Big Data

Over the last couple of years, 'Big Data' has become a prominent theme in discourses in the marketing community. We have access to more data, and more types of data than ever before. The costs of acquisition, storage, dissemination and processing of data are ever diminishing.

While these factors make it possible for a marketer to engage in data-driven marketing, increasing local and global competition for the customers' 'share of wallet' and 'share of mind' makes it necessary for a marketer to rely on data-driven insights. Dependence on data is not limited to isolated campaigns, but data-driven marketing is a different paradigm altogether.

It all begins with understanding your customer better. The focus of this exercise is to have an answer to the following questions:

Who are my customers? Which of their other needs could I fulfil? How can I best communicate this fact to them?

In an ideal situation, one would be able to get a 360o view of the customer, combining data from various sources:

Profile data: Age and gender at the very least. Other nice-to-haves are family status, income, ethnicity, language, vehicle and consumer durable ownership and acquisition channel.

Contact details: Their physical address, e-mail address, telephone number- so as to be able to contact them. One could use this data to impute things like area of residence, city, e-mail domain etc.

The transaction is the basis for the customer's association with the marketer's brand. This is the place to look, if one wants to understand the customer's fulfilled and unfulfilled needs/wants. Starting from the complete, detailed transaction history, one can easily compute things like tenure of association, recency, frequency, current monetary value. Some of the other interesting attributes are entry product (what drew them to me in the first place), favourite product/product category, time between purchases, timing of purchase- time of day/day of week, any indicators of seasonality, discount seeking behaviour and so on (the list is practically endless). Moreover, one could use the transaction data to impute the customer's lifestyle, life-stage or changes in life-stage. A recent story that made the headlines was about the U.S. based retailer Target being able to impute their customer being in pregnancy.

Campaign and response data: For all practical purposes, this data is equally (or more) important to the marketer as the transaction data. As we discuss later, a crucial element of the data-driven marketing paradigm is to continually 'test and learn'. The campaign and response data is what allows the marketer to do so. One can use this data to answer questions such as: What combination of media/time of campaign/offer work the best? What is the immediate and long term impact of a particular type of campaign? Depending on the technology, e-mail campaigns can also tell you what browser/OS were used to access the e-mail.

Web data: If the web browsing data can be tagged with a unique customer identifier, it can be a source of great insights. Some of the interesting variables here are products, product categories browsed, journey within the site, and entry and exit points. While many companies analyse weblog data in isolation, the most rewarding insights can be obtained if web data can be married to transaction, profile and campaign data. For a client that was able to achieve this linkage, the browsing variables were always among the top predictors of purchase or campaign response.

Social media data: Most businesses have a social media presence along with a number of people who follow/like them. The key to usefulness is again linkage with other data sources. If that can be achieved, some interesting data points can be extracted from social media such as sentiment and influencer status.

One would use all of the above data points to get a holistic view of each and every customer. This view can be put to various uses, some of which are as follows:

Segmentation: This involves strategic long term segmentation as well as tactical topical segmentation. The strategic segmentation is generally used to track movement (or lack of it) across segments, while the tactical segmentation is generally used for campaigns.

Predictive modelling: The rich customer view can be used for efficient targeting by predicting the propensity for purchase, campaign response, churn, cross-sell/up-sell etc.

Profiling: The data and insights driven approach allows a marketer to design and fine tune his offer, communication, and even the creative based on profile of a customer who exhibits the desired behaviour.

Cross-sell opportunities: The association between products can be used for innovative design of campaigns. For example, customers who buy Whiskas cat food are more likely to be interested in olive oil as compared to a regular customer. Product associations can be judiciously used for fine-tuning the campaign offer, communication and creative.

Summing up, in an ideal world, the data enrichment process enables the marketer to find an answer for the questions: "Who are my customers? Which of their other needs could I fulfil? How can I best communicate this fact to them?"

The key requirement is the availability of data across the board and the ability to achieve linkages between them. Reality, as we all know, is rarely so straightforward. To circumvent this challenge, the aforementioned 'test and learn' strategy is the marketer's best friend.

The data driven marketing paradigm comprises of 3 phases:



Identify and measure: Identify attributes critical to the business and measure the current levels. The attributes may be at a systemic level or at an individual customer level. Measurement and continuous monitoring of the attributes allow the marketer to evaluate success and formulate the 'test and learn' strategy.

Test and learn: in this phase one puts to test the assumptions, hypotheses and hunches. The best way to achieve this is through systematically designed campaigns. In addition to generating ROI, the campaigns are a great practical way for the marketer to connect the dots (or at the very least create more dots to be connected) Improvise and improve: The learnings from the campaigns need to be institutionalised, while you continue to be on the lookout for changes in them. Strategies that worked earlier need not continue to do so. Continuous monitoring of the numbers and trends will enable you to pick up some early indicators, and respond to those.