Case Study
Commercial Analytics
Market Basket Analysis
Which products are purchased together, and how can those associations be used responsibly?
Download PDF
A complementary visual resource for a deeper view of the case.
The business challenge
Which product combinations occur together with meaningful frequency, and which commercial hypotheses are worth testing?
Understanding baskets can inform merchandising, recommendations and promotions. Popularity, availability and seasonality can explain patterns without a causal relationship between products.
Key business questions
- Which products appear together more often than expected?
- Which patterns remain stable by channel, store or period?
- Which associations warrant a commercial test?
Data & evidence
- Receipts or orders with consistent identifiers.
- Date, store, channel, promotion and availability.
- Product hierarchy and attributes.
- Relevant returns and cancellations.
How I would approach it
- Define the basket, time window and inclusion rules.
- Explore penetration and basket size.
- Calculate support, confidence and lift.
- Review stability by period, channel and store.
What you would get
- Prioritised map of product associations.
- Interpretable rules with volume and relevance context.
- Hypotheses for recommendations, merchandising or bundles.
- Incremental validation plan.
From analysis to action
Illustrative example: High lift indicates co-occurrence above that expected under independence; it does not prove that promoting one product causes purchase of another.
- Prioritise combinations for experiments.
- Review assortment or navigation.
- Design bundles with margin guardrails.
How success should be measured
- Rule coverage and stability.
- Conversion and incremental value in tests.
- Margin and substitution effects.
Other approaches to consider
Product embeddings, sequence models or networks can complement association rules where sufficient scale exists.