Using Market Basket Analysis ... is it sufficient?

By Hari Prasad Kafley

Market Basket Analysis based on affinity algorithms is one among the most popular and commonly used techniques to analyze market basket data mostly applicable to consumer package goods. The association rules tend to identify a basket of products which are likely to be bought together. With increasing complexity and competitiveness in consumer package goods industry due to information revolution, mass merchandising is on decline. Consumers are increasingly recognized into large number of distinct segments. Diverse items are available for sale. Products category contain hundreds of competing products. High degree of differentiation is being achieved leading to strong brand names offering strong competitive advantage. Temporary price reductions, temporary product assortments and placement at stores are not sufficient for inducing customer loyalty. Due to greater degree of information access, consumers are highly mobile and knowledgeable to make purchase decisions based on price and differentiation of products. Eventually, retailers having greater ability to negotiate favorable terms with suppliers and manufacturers reap the advantage.

With maturity of consumer package goods industry, mere opening of new stores to sell to larger number customers hardly ensures profitability. Essentially, it important to retain existing customers with increased product portfolio. Retailers those are able to reduce their operating costs, manufacturers achieving economies of scale increase profitability. Operational efficiency can typically increased with introduction of customer loyalty program leveraging on information technology.

Present day demands great deal of collaboration among retailers, suppliers and manufacturers. Actionable recommendations from analysis should cater to the needs of all partners in the chain. Market Basket Analysis tends to give biased recommendations profitable to retailers. Though statistically the results are credible but in this competitive economic environment and informational open market results may not translate into actionable strategy.

Problem takes origin from affinity algorithms in Market Basket Analysis which takes into account individual and isolated rules. Current need is to generate rules to predict the association among group of products. Complex algorithms being practiced can be used to include product categories, but it leads to increase in rules exponentially. The rules appear to be complex leading to difficulty in interpreting the rules. Furthermore, affinity algorithm in itself does not assume causal assumptions whereas the output of the algorithm suggests causal relationship among the products. Hence there is a need to introduce either a new and causal form of affinity analysis.Extension of algorithm to produce rules to be applicable across larger groups of transactions needs to be explored.