Market Basket Analysis (MBA) is a data mining technique rooted in association rule learning that identifies products frequently purchased together. Originally developed for brick-and-mortar retail (analyzing what's in a shopper's "basket"), it's now a cornerstone of modern e-commerce analytics.
How it works:
MBA analyzes your transactional data to find statistically significant product associations. For example, if 68% of customers who buy a "Classic Cap" also buy a "Logo T-Shirt," that's an association rule worth acting on.
Key metrics in MBA:
- Support: How frequently the combination appears in all transactions
- Confidence: Given product A is purchased, the probability product B is also purchased
- Lift: How much more likely products are purchased together compared to independently
E-commerce applications:
- 1Product bundling: Create bundles from high-confidence associations
- 2Cross-sell recommendations: "Frequently bought together" sections
- 3Store layout optimization: Position associated products near each other
- 4Inventory planning: Stock associated products together to prevent missed sales
The Apriori algorithm is the most common approach to MBA, though newer methods like FP-Growth offer better performance on large datasets. The quality of your MBA depends entirely on the quality and volume of your transactional data — typically you need at least several hundred orders to find statistically meaningful patterns.
