Market basket analysis: what baskets show that totals don't

A shopper sets three items on the counter: a cold drink, a bag of chips, and a phone charger. The register total says the store made a sale. The combination says something more useful. Market basket analysis is the study of which items are purchased together in a single transaction, and of what those pairings reveal about how people actually shop.

Sales totals tell you what sold. Baskets tell you what sold together, and together is where the strategy lives.

What is a market basket, exactly?

A basket is the full contents of one transaction: everything a shopper bought in a single visit, captured on one receipt. In scan data terms, it's the set of line items sharing a transaction record. A basket can be one item or twenty, and its size and makeup are data in their own right.

The distinction from ordinary sales reporting matters. Aggregate reporting treats each item's sales as its own column. Basket analysis keeps the transaction intact, preserving the relationships between items that column-by-column reporting throws away.

Analysts describe those relationships with affinity measures, which compare how often two items appear together against how often they'd coincide by chance. The math has formal names, but the idea stays plain: is this pairing a habit or an accident?

What do co-purchase patterns reveal?

Occasions, mostly. When two items appear together far more often than chance would suggest, there's usually a shopping mission connecting them: the morning stop, the lunch run, the evening errand. Reading those missions from the data tells you what job the store is doing for its customers at different hours.

The patterns have direct commercial uses. Placement, because items that share baskets are candidates for shelving near each other. Promotion design, because a deal on one item can lift its frequent partners without discounting them. And assortment, because an item that anchors many baskets is worth more to the store than its own sales line suggests.

One caution: co-purchase is correlation. The data shows that items travel together, not that one causes the other to sell. Treat basket patterns as leads to test, not verdicts.

A hypothetical shows the practical value. Suppose a brand's iced tea turns up unusually often in baskets with a particular salty snack. That's a testable idea within the month: pair them on the counter card, or place them within arm's reach of each other, and watch whether combined velocity moves. The basket suggested the experiment; the register grades it.

How is basket analysis used in small-format retail?

Small stores make baskets especially interesting because trips are short and purposeful. A basket in a bodega or convenience store often maps to a single mission, which makes the pairings cleaner to read than in a supermarket cart holding a week of groceries.

For a consumer packaged goods (CPG) brand, small-format baskets can answer practical questions. What does my product get bought with? Does it anchor the trip or ride along? Those answers shape counter placement conversations, multi-item promotions, and even pack-size choices. Basket size itself is a signal worth watching: a store whose average basket grows is doing a bigger job for its shoppers, and an item that appears in growing baskets is part of that story rather than a bystander. Any single store's baskets are a small sample, so the useful reads come from aggregating across many stores, which is what a network-level dataset such as the one behind NRS Insights makes possible.