Shivaraj Chabungbam
Student, PGDM 2014-16

With the rapid adoption of technology in business enterprises, we are witnessing two big trends emerging today: the continuous digitization of business processes, where manual processes are being replaced by automated ones and the increasing use of technology to aid in making business decisions. The former is enabling enterprises to generate massive amounts of data from their automated processes, which can be mined and the latter, enables companies to use technology to gather deep insights about their businesses. With the decreasing cost of storing large amounts of data and increasing computing power of IT systems, enterprises are now able to employ complex algorithms to extract large amounts of actionable information from the data.

Big data analytics is the term used to refer to this processing of large amounts of data for extracting actionable insights. Enterprises with big data analytics platform in place are now showing an insatiable appetite for data and this trend of using data-driven decision making is catching up fast in most industries as more and more companies are trying to incorporate analytics into their processes. This paper attempts to study how the revolution in big data analytics has impacted the retail industry and the possible ways in which this technology can change how retail companies do their business.

Rise of Omni-channel marketing strategies

About 20 years ago, as the Web became open for commerce, the internet became an important vehicle for consumer purchases. A major fillip was given to this new development by the rise of online retailing giants such as Amazon and eBay. and was first launched in 1995 with the former, as an online shopping site for books. Business was classified on the basis of their presence online or offline. However, the current trend is about having an omni-channel presence which means companies not only are trying to have both an online and offline presence, they are also trying to create a consistent user experience whether a customer buys online or offline.

One such capability required by a company to enable this is to understand the customer preferences. Analysing customer preferences is filled with ambiguity because not all customer preference combinations are available at different price points. However, using various analytical techniques and customer preference data (Stated preference data) a better model for measurement of customer preference can be built. Social media too provides a valuable ground for the collection of customer feedback and understanding customer preferences.

Data on customer preferences gives rise to opportunities for understanding purchasing behaviours, giving better product recommendations and upselling or cross-selling.  When Target statistician Andrew Pole built a data mining algorithm which ran test after test analyzing the data, useful patterns emerged which showed that consumers as a whole exhibit similar purchase behaviors. It got a little out of hand, when Target accurately predicted that a teen girl was pregnant (even before her family knew) and sent her customized products’ catalogue to ease her buying needs. Amazon has pioneered the personalization strategy by using customer data to provide product recommendations to customers based on previous purchase history, browser cookies and wishlist.

Data on consumers also allow for better customer segmentation so companies can target the high value customers which will increase their ROI in marketing. Customer Analytics can be classified into descriptive, predictive and prescriptive analytics. Descriptive analytics looks at past performance and understands that performance by mining historical data to look for the reasons behind past success or failure. Almost all management reporting such as sales, marketing, operations, and finance, uses this type of postmortem analysis. Such traditional descriptive methods (commonly seen in datawarehousing systems) are powerful for describing the population but are limited in their predictive abilities with respect to the same population. Predictive models select defining attributes differently- ie. while we may know a lot about our customers, we may not be able to accurately forecast what they will do next. In these instances applications of predictive models might help construct a list of customers for a survey that will be more accurately target its lift ratio when applied to the real population. Prescriptive models however, take advantage of the data of descriptive models and the hypothesis of predictive models and try to answer, not only what the customer will do next, but why they will do so. This sort of lateral analysis opens up inquiries into questions not anticipated earlier in the design cycle, but need to be asked as the rigorous changing real-time needs of the dynamic customer system. Techniques such as Look-alike modelling can identify customers who share similar features with those customers whose behaviour can be predicted with statistical rigor.

Rise of better predictive models

With the depth of insight on customer behaviour, companies are also now in a position to handle credit card fraud during purchase. Amazon has an intensive program to detect and prevent credit card frauds, which has led to 50% reduction in frauds within first 6 months. Amazon developed fraud detection tools that use scoring approach in predictive analysis. This retail analytics depends on huge datasets that contain not just financial information of the transactions but it also keeps a track of browser information, IP address of the users and any other related technical data that might help Amazon refine their Analytic models to detect and prevent fraudulent activities.

Even in domains such as supply chain management, big data has a played a pivotal role in increasing the efficiency of the processes. It allows for real-time tracking of shipments, inventory optimisation, forecasting. Metro Group retailer uses retail analytics to detect the movement of goods inside the stores and display relevant information to the store personnel and customer. For example, if a consumer takes an item into the trial room, the product recommendation system recommends other related products while the customer is trying on the apparel. The store personnel inform the customers whether the products are in sock or not. The retail analytics system of Metro Group also keeps a track of the movement patterns on and off the shelf for customer analytics for a later point in time. Retail Analytics also alert managers at Metro Group about abnormalities in product by identifying unusual patterns, for example the product is taken off from the shelf several times but it is not purchased.

Pricing intelligence using big data analytics generates higher revenue for companies. With so many customers comparing online and offline prices, it can generate better conversion rate and revenue by indicating the right price to be charged to the customer. Such a system compares the prices offered by various competitors in real-time and offers the best price to the customer. Sometimes when a customer is shopping in a retail store, he/she is willing to overlook a small difference in price in favour of instant gratification thereby giving room for higher margins through better price offering. Amazon’s analytical platform has a great advantage in dynamic pricing as it responds to the competitive market rapidly by changing the prices of its products every 2 minutes (if required) whilst other retailers change the prices of the products every 3 months. is another giant retailer leveraging retail analytics for dynamic pricing by identifying various opportunities for price optimization to generate incremental revenue and margin. Other retailers that leverage Retail Big Data Analytics are RadioShack and Groupon.

By any measure, it can be seen that big data analytics is about to revolutionize the retail industry in the way business was traditionally conducted. As more and more retailers begin to adopt these strategies, customers will see a paradigm shift in their shopping experiences.