Market Basket Analysis and Frequently Bought Together

Association-rule mining looks for item combinations that appear together in transaction data. For a rule X to Y, support is the share of all baskets containing both X and Y. Confidence is the share of baskets with X that also contain Y. Lift compares that confidence with the overall frequency of Y; lift above one means Y appears more often with X than its baseline rate in the analyzed data. These measures answer different questions. A rule can have high confidence because Y is popular overall, while a high lift based on very few transactions may be unstable. Direction matters: X to Y and Y to X can have different confidence because the base rates differ, even though support and lift for a two-item pair are symmetric. Rules describe co-occurrence, not causation, preference, or an ideal product bundle. Promotions, seasonality, store layout, and customer segments may affect the pattern. A recommendation widget can improve discovery, but teams should test customer outcomes and guard against exposing sensitive inferences. Data quality matters: transaction boundaries, returns, and time windows change results. Analysts should report counts and periods, check whether patterns replicate, and avoid targeting customers with assumptions unsupported by the data. Basket analysis is a useful descriptive tool when interpreted with context and tested before operational use. A rule discovered after many searches can arise by chance, so analysts should confirm it on later transactions before changing merchandising.