Market Basket Analysis – The other MBA

  • 15 Dec 2015
  • Posted by Sashindar Rajas...

Meet Jim. Jim is looking to throw the ultimate birthday party for his 4 year old son and has quite a ‘to- do’ list for the party. Jim decides to kick off his party prep by logging on to his trusty ecommerce store and begins shopping, beginning with ingredients for the birthday cake. He spends a few minutes filling his cart with essentials like flour, eggs, milk, sugar, frosting and candles. Done with his purchase, Jim is on the verge of the checkout. But just before his payment screen, Jim encounters an offer.  Apparently, the same store stocks Party Hats, Paper Plates, Goody Bags and even offers bookings for a children’s entertainer. All offered as a package with a reasonable discount. Jim is relieved that his job just became a lot easier and is pleased that his store is so thoughtful about his needs. Happy Jim adds more items to his cart and subsequently finishes his entire preparation in a single go.

Now we all know what Jim had a huge party to plan and just set out to buy what he needed for a cake. But how did the store in a few short minutes arrive at the same conclusion and so seamlessly made his life a whole lot easier? Through Market Basket Analysis, the store was able to identify what Jim was buying, conclusively decipher his intentions and intelligently offer what he was looking for. As a result, the store is left with a sizeable increase in order value and a happy customer who now knows where to go to for his party needs. So how did Market Basket Analysis make everyone a winner that day?

Market Basket Analysis works by identifying items that are most likely to be purchased together when a customer shops online.  It determines the intrinsic relationships and affinity between two or more products by analyzing how frequently the combinations of products occur in a given set of purchases. Using this logic any ecommerce marketer who understands and has the support for  Affinity analysis & Association rules / Apriori algorithms can then determine the best combination of products to push to customers at the right time, thereby maximizing the chances of purchase.

While some might argue that identifying complementary products is easy enough, the real value of Market Basket Analysis is felt in an ecommerce environment carrying a few hundred different product lines and when identifying relationships between the seemingly unconnected product lines.

 In Jim’s case, his initial individual purchases were grocery. Together his purchases could form the basis for any baked good like cookies, or pies. But throw in a candle as a purchase and the basis for a cake becomes more likely. And where there’s a cake there’s a high likelihood of a party which leads to the store to push offers on party supplies to Jim. These two categories are as different as chalk and cheese, but under the lens of Market Basket Analysis, combining them makes absolutely perfect sense.

Market Basket Magic

Market Basket Analysis works by generating a set of rules that govern how linked various products are. This can be likened similar to an equation with values either side, with the left side containing the products purchased by the customer and the right hand side carrying the products that might very likely be of interest to the same customer. Consider Jim’s example. In his case the rule would look something like,

IF {milk, flour, sugar, eggs, candles} THEN {party hats, paper plates, magician}

Market Basket Analysis can generate these sets of rules which when subject to further analysis can be leveraged to undertake data backed marketing decisions.

While Market Basket Analysis outputs are key, the Rules are usually be further scrutinized and chosen for best performance. Keeping the mathematics minimal, some criteria that decide the validity of Market Basket Analysis’ Rules are as follows.

Support: The Support of a product or group of products denotes the number of times the product or set of products appears in a given set of transactions expressed as a ratio. The accepted rule of the thumb is that the higher the support in a rule the better.  

Confidence: The probability that denotes that a customer purchasing a single or a group of specific products is then going to purchase the associated products identified by the rule. Confidence is a key factor in deciding the worthiness of the rule and is also expected to be higher.

Lift: Another ratio that expresses if and how the presence (purchase) of certain products by a customer influences the purchase of its associated products. A Lift ratio greater than 1 indicates an increased probability that the purchased product as well as its associated product will feature in a transaction. Similarly a Lift ratio less than 1 indicates a decreased probability of the associated product occurring in a transaction. An exact value of 1 denotes the products and its associated products are independent of each other. Lift is quite often referred to define the strength of a Rule.

Market Basket Analysis in the real world

While the concept behind the Market Basket Analysis is simple enough, the crux of the matter lies in implementing it in a contemporary ecommerce environment.  With a store carrying a few hundred product categories and a few thousand products with a rotating inventory makes performing a manual Market Basket Analysis quite simply, inefficient. However there exist a plethora of algorithms and tools that relatively simplify the process. While standard data mining tools such as PolyAnalyst, R can be effective for Market Basket Analysis, the need for a certain degree of statistical and operational expertise is inevitable.  This is where specialist tools such as Greyferret can help the average ecommerce marketer get the benefits of Market Basket Analysis while eschewing complexity. In addition to performing the analysis, tools like Greyferret can interpret the results, offer excellent insights into customer behavior and aid in effective decision making.

What Market Basket Analysis means to the ecommerce Marketer

Market Basket Analysis opens up new avenues for ecommerce marketers and can lead to a sustainable revenue stream if leveraged correctly. Jim’s birthday shopping encounter would have generated a bulk of data pertaining to his purchase behavior such as products viewed, brand preferences, price points, delivery details and so on. The wealth of this data can help the store engage with Jim to a greater degree. For example,

  • Analyzing Jim’s purchases over time, the store can send out coupons and special offers on products that might interest him via email and SMS
  • Provide a personalized web experience with a personalized welcome message, a list of his frequently brought products, pre-selected payment methods, delivery address and so on
  • Present recommended and complementary products during his next shopping experience
  • Help increase retention and loyalty. The store can remember Jim’s party purchases and towards the next birthday occasion, send him an automated alert with the latest products in store. Such as positive engagement will most likely retain Jim as a long standing customer

Undoubtedly, the key to the success of the above requires multiple critical systems. From a robust ecommerce platform to collect the data, a comprehensive analytics suite to interpret the data and an omni-channel marketing platform, seamless workability is vital to ensure mutual benefits for stores and customers. Fortunately, multi-level integration is fast becoming the norm in most ecommerce architectures.

Tools like Greyferret provide deep and actionable insights into customer behavior, while integrating with popular ecommerce platforms (such as Shopify, Magento and Bigcommerce) and also with a wide variety of enterprise level marketing platforms (Mailchimp to Responsys). This can greatly reduce the complexity of the entire ecommerce architecture and streamline the marketing process by eliminating the need to switch back and forth between datasets and tools.

 As ecommerce becomes increasingly dependent on data driven practices it is inevitable that marketing follows suit. We are not in an era where that the availability of comprehensive data is no longer a challenge, but in an age where the understanding and effective utilization of the data is. So as marketers, go ahead and take the plunge. Give Market Basket Analysis a try and pioneer a transformation in your ecommerce store.

To find out more on the tools and process which can help you get started using Market Basket Analysis, visit