BigQuery adds market basket questions to conversational analytics

What changed

Google's 3 September 2026 BigQuery release note marks conversational questions about market basket analysis as generally available. The accompanying documentation excerpt explains that the task looks for products frequently bought together in a transaction. It describes built-in query templates and association measures, giving analysts a conversational route into a familiar retail-data question.

The advertiser opportunity is to turn purchase patterns into testable bundle or cross-sell ideas. For example, a retailer could investigate whether buyers of a particular accessory often purchase a related main product in the same order. That pattern could justify a proposed bundle creative or landing-page test. It would not, on its own, establish that showing the accessory in an ad will increase revenue. Popular items can appear together simply because each is bought often, and an existing promotion can influence what looks like natural demand.

The first useful work is therefore to define the dataset and the question. An order-line table needs a reliable way to identify which items belong to the same transaction. The period, treatment of returns, product variants and promotional bundles also need to match the decision being considered. A conversational answer cannot repair a misleading commercial definition. Ask the analyst to check a sample of the underlying orders before a promising-looking association becomes a creative brief.

Why it matters for advertisers

This can make exploratory retail analysis more accessible to media teams that know the commercial question but need help expressing the query. The value is a shorter path to a defensible hypothesis, followed by a campaign experiment that measures the business effect. Keep the basket association and the result of the advertising test separate in reporting so that an exploratory correlation is not presented as incremental sales.