“IF … THEN …” is a conditional statement that is broadly used in programming languages. We use the same idea in Market Basket Analysis to analyze the purchase behavior and find patterns based on the theory that if you buy a particular group of items, it is more (or less) likely to buy another group of items.
Maybe you were waiting at the counter in some grocery store and bought a pack of chewing gum or a candy bar. Or perhaps on a trip to a convenience store to buy soap. It is not coincident that gums and candies or shampoos and conditioners are stocked near each other. Usually, these recommendation systems and the ways items are stocked are based on the subjective experience of marketing professionals and inventory managers.
With the advancement of data mining and machine learning, techniques could be used on many transactional data to uncover hidden and interesting patterns in purchasing behavior. Market Basket Analysis is used to find associations in large databases using statistical performance measures. The information revealed from Market Basket Analysis helps retailers to manage better categories by understanding buyers’ needs and developing cross-promotional programs and occasion-based consumption insights, or even capturing new buyers.
The result of Market Basket Analysis is a set of association rules that specify patterns of relationships between items. A typical rule might be expressed in the form: IF {Eggs, Butter, Baking Powder} THEN {Balloons}, this is a type of rule that is based on the consumption occasion which indicates that buyers usually buy these items when they are preparing a cake for a birthday party.
Although association rules are used in market basket analysis and supermarket shoppers, they are also instrumental in finding patterns in many other data types. For example, analyzing credit card purchases, identifying fraudulent medical insurance claims, recommending books in online stores based on previous preference, recommending movies or series based on past viewing behavior, etc.
Sapience has been recently working on association rules and algorithms that are strong in tackling large-scale transnational data and producing results that are easy to understand and implement.
Read more about our analytics solutions