Retailing today is nothing if not a hypercompetitive industry. A complex & rapidly changing landscape, stiff competition, and ever more demanding customers are pushing progressive retailers to rethink how they operate. Their responses have included strategies of scale (consolidation, growth, international expansion, etc.), innovation (alternative store formats, brand extensions, etc.) as well as promotion (both online and offline), among others.
Irrespective of grand strategies and visionary shifts across the globe, though, it is a truism that the need to dive deep and understand the consumer mindset continues to sit at the center of all that a retailer must do. And over the last couple of decades, technology has played a tremendous role in informing strategies and tactics down to the SKU level.
Probably the most critical way that technology has aided the retailer is by enabling the capture of large volumes of consumer transactional data at very reasonable costs. Retailers can now obtain terabytes of information about their customers’ buying patterns, demographic information as well as (through various means) psychographic insight. This information can answer important questions including: When did the customer shop? How was the payment made? How many and what specific items were purchased? What was the relationship among the purchased items?
There is no doubt that this vast point-of-sales (POS) data has (when effectively utilized) empowered the retailer to better understand their business and improve decision making. Proactive retailers use this information to deliver targeted offerings that are aligned with consumer expectations and subsequently deliver positive revenue impact.
That said, though, how do retailers tactically use these terabytes of information?
Many times, as consumers, we tend to overlook how goods are physically arranged in a grocery store or supermarket. What might look (to us) to be ‘random distribution’, is actually a meticulously planned arrangement of goods. At its analytical core, the grocer assesses the purchasing pattern of his/her customers and arranges these purchased products accordingly.
Simply stated, every customer’s basket tells a story. The key is to discern his/her preferences – preferences that are buried deep inside the shopping basket. This technique of discovering relationships between products purchased together is known as Market Basket Analysis (MBA). As the name suggests, MBA essentially involves using consumer transactional data to study buying patterns and exploring the possibilities (and probabilities) of cross-selling. The objective of MBA is to utilize sales data effectively to improve marketing and sales tactics at the store level.
Since shelf space is limited, retailers have significant incentives to make these placement decisions correctly. Indeed, product assortment, product display area selection, shelf space allocation, and inventory control are critical retailing operations that directly impact the financial performance of retail stores. Thus, typical insights that can be derived from MBA are:
Table 1: Illustrative Purchases made at a Grocery on a given day
- Products X,Y are typically purchased together
- If products X,Y are purchased then product Z may also be purchased
The applications (and financial value) of understanding insights such as the above are evident.
The Mathematics of MBA
MBA works on the concept of probability of co-occurrence of events. There are several specific metrics that are taken into consideration to assess how strong the co-occurrence is:
- Support: Ratio of the number of transactions that includes both Product A & Product B, to the total number of all transactions
- Confidence: Ratio of the number of transactions that include both Product A & Product B, to the number of transactions that include a particular product
- Lift: How good the rule is at predicting the “result” or “consequence” as compared to having no rule at all (i.e. to what extent is one better of deploying the rule rather than simply making a guess).
Let’s consider a simple example. Suppose that within a particular grocery store, a total of 80 transactions take place. Furthermore, two products – bread and butter – are purchased 60 times together out of this total of 80 transactions (refer to Table 1 above).
Based on this data, the probability of bread and butter being purchased together is 75% (i.e. 60 occurrences out of 80 in total). This is technically referred to as the support of buying bread and butter.
In addition, if there were 75 transactions (in total) for bread, then the confidence of buying butter along with bread is 80% (60 occurrences where bread and butter were purchased out of a total of 75 transactions involving bread).
Further, if we assume there were 60 transactions for butter, then the lift of buying butter with bread is 1.06 ((60 occurrences for combined purchase of bread and butter X 80 total occurrences)/ (60 transactions for butter X 75 transactions for bread)).
Table 2: The Confidence of the same Purchases made at the Grocery
Indicates that there is 80% confidence of buying butter if bread is purchased
Statistically, high support, confidence and lift are signs of solid associations between the occurrences of the events. Even demographics can be incorporated and utilized in developing association rules as part of MBA. For example, consumers in the age group of 8-15 years may have a higher propensity for purchasing video games than those in any other age groups (Here age group is variable A and video games is variable B).
The application of MBA can be found in non-retail sectors too. MBA has been used to develop decision tools for traffic safety. Specifically, a study had been conducted in Florida to analyze the patterns of crashes. Crashes were analyzed in conjunction with a range of other variables and the interdependence among these crash characteristics was studied. The analysis found significant dependence of the severity of crashes on lack of illumination. Further, it was found that under rainy conditions, straight sections with vertical curves were crash prone.
MBA: Benefits galore
The examples above, while simplistic, illustrate the power and value of a tool such as MBA. Indeed, its application is not limited to grocery stores alone but to other retail (and non-retail) subsector in operation today. Furthermore, MBA can be applied to online purchases equally as well. We see examples of this all the time at online stalwarts such as Amazon. Their recommendation to buy products are based on a careful analysis of the previous purchase history of consumers along with the purchasing history of those that bought any specific product.
Indeed, retailers adopting MBA are able to ensure that their product offerings and promotions match as closely as possible to consumer expectations, ultimately providing targeted campaigns and segment specific offers that generate substantially higher returns.
Author: Nitin Jain, Strategic Services Practice