Market Basket Analysis: Support, Confidence, Lift & Cross-Selling

What you'll build

What you will be able to do

Each lab ends with a concrete result that you build yourself and the checker confirms, cell by cell.

• Turn 24 shopping baskets into item counts, support values and the average basket size, the starting point of every market basket analysis.
• Count how many baskets contain two items at once with SUMPRODUCT, fill the whole item-by-item co-occurrence matrix with one formula and turn a pair count into a pair support.
• Look up item and pair counts in the co-occurrence matrix with INDEX and MATCH, calculate the confidence of eight rules and see why Bread → Butter and Butter → Bread are not the same rule.
• Calculate support, confidence and lift for ten rules from a month of baskets and label every rule as complement, independent or substitute.
• Add two more rule metrics: leverage, which measures how many baskets a rule adds over chance, and conviction, which measures how much the rule depends on its direction.
• Apply minimum support and confidence to 14 candidate rules, count the survivors, rank them by lift and pull out the strongest rule, and see why the rule with the biggest lift drops out.
• Flag the frequent pairs, use them to keep or prune five candidate 3-item sets the way the Apriori algorithm does, and count the support of the candidates straight from the baskets.
• Turn five rules into a cross-selling plan: how many baskets can still receive the recommendation, how many extra units it sells, what that is worth after the coupon, and which rule to roll out first.