Basket Composition: Reading Product Mix Without Overstating Intent

Two transactions can have the same item count and value while telling different commercial stories. One may contain a complete family dinner across several categories. Another may contain six units of one promoted drink. Basket size sees a tie. Basket composition sees the difference.

Basket composition describes how a transaction is structured across categories, brands, price tiers, pack types and product roles. It can help retail media teams understand the context around a purchase, compare exposed and control baskets, and identify how a campaign changed the mix.

It does not reveal intent with certainty. A basket is behavioral evidence, not a transcript of the shopper’s reasoning.

What basket composition includes

A composition profile can contain:

• category and subcategory shares;
• brand and private-label mix;
• premium, mainstream and value tiers;
• promoted and non-promoted products;
• units, distinct SKUs and category breadth;
• fresh, ambient, frozen or other operational groups;
• target, substitute and complementary products;
• observed transaction labels such as promoted, repeated, first-time, substitute candidate, complement candidate or unresolved.

The profile can be expressed by item count, quantity, revenue or margin. State which base is used. A category can represent a small share of units but a large share of value. Shares are compositional and sum to a fixed total: one share can rise because its numerator grew, because another part of the basket declined or because both happened. Read every share beside absolute units, value and containing-basket counts.

Google Analytics ecommerce documentation supports item-level fields such as identifier, brand, category, price and quantity across cart and purchase events.[3] Retailers need consistent taxonomy and product attributes before those fields can support a reliable composition profile.

Whole-basket composition differs from directional pair analysis and from recommendations made while a basket is still open. It profiles the completed transaction rather than answering either adjacent question.

Build the profile in layers

Transaction layer

Start with the purchased basket. Count units, unique products, categories and transaction value. Record promotions, refunds and substitutions.

Product-role layer

Keep observed attributes separate from inferred roles. Prefer transaction labels supported directly by events, such as promoted, repeated, first-time, substitute candidate or complement candidate. If a role such as routine, treat or destination is used, document the rule, confidence, permitted use and reason shown to an analyst. Provide an unresolved fallback rather than forcing a mental-state label. Roles can change by mission, so never make them permanent shopper traits.

Shopper and trip layer

Add channel, store format, time, cadence and mission proxies. These variables help explain why composition differs, but they should not be confused with product composition itself.

Commercial layer

Identify the advertised product, competitive set, promoted tier and campaign eligibility. This connects the basket profile to the decision the media plan is supposed to influence.

A useful composition table keeps each layer separate. Blending every feature into one segment may produce a label that no buyer can explain or activate. Freeze or version the product hierarchy for the analysis window, map reclassified SKUs consistently across pre and post periods, and disclose products that are unmapped or assigned to several categories. Otherwise taxonomy drift can manufacture a composition change.

Bought together does not always mean one intention

Within-basket recommendation research notes that baskets can contain multiple intentions and that shoppers can switch between them during a session.[4] A grocery basket may combine dinner, breakfast and household replenishment. A marketplace order may combine unrelated needs because delivery economics favor one checkout.

Research on complementary products also treats basket and browsing behavior as evidence from which relationships are inferred, not as a perfect statement of intent.[5] A recent industry preprint warns that raw co-purchase data can generate noisy or inappropriate complement recommendations without constraints and review.[6]

Use cautious language:

• “associated with” rather than “caused by”;
• “consistent with a meal mission” rather than “the shopper intended a meal”;
• “candidate complement” rather than “required product”;
• “observed in the same basket” rather than “preferred together.”

This discipline improves both privacy and decision quality.

Composition patterns that can improve media planning

Category breadth

A broad basket may support occasion or stock-up creative, while a narrow basket may need a direct category message. Breadth is context, not automatic permission for more ads.

Brand portfolio mix

A multi-brand company can see whether the advertised brand appears with other portfolio products, competitors or private label. The result can shape cross-brand planning, but shared ownership does not prove a useful shopper relationship.

Price-tier mix

The target product may enter premium baskets, value baskets or a mixed-price mission. This can guide creative and assortment decisions. Do not infer household income from one transaction.

Promotion mix

A campaign-period basket may become more promotional even when value rises. Separate media, price and promotion effects.

Complement and substitute mix

The promoted product may add a new role, replace a competing product or shift pack size. Each effect has a different commercial meaning.

For adjacent-category relevance, category affinity can help define audiences. Basket composition then tests what actually appeared in the transaction context.

A worked example

A premium coffee brand runs onsite and in-store media. Sales and average basket value rise among exposed shoppers.

Composition analysis asks what changed inside those baskets. Did premium coffee replace mainstream coffee? Did shoppers add biscuits or breakfast products? Did they buy larger packs because of promotion? Did the exposed group simply contain more weekly stock-up trips?

The report can compare:

• coffee-category incidence;
• brand and tier share within coffee baskets;
• category breadth and units;
• complementary breakfast-category presence;
• promoted-item share;
• total value and margin;
• treatment and control mission mix.

If premium brand share rises while coffee-category incidence stays flat, the campaign may have shifted choice within the category. If category incidence and breakfast breadth rise together, the campaign may have expanded the occasion. If only promoted pack size changes, price mechanics may explain much of the result.

Composition turns one sales line into several testable mechanisms.

Compare like with like

The IAB/MRC guidelines emphasize consistent definitions, transparent data sources and clear attribution scope.[1] Composition comparisons should use the same product taxonomy, transaction rules, period logic and availability filters.

At minimum, compare:

1. Pre-treatment balance: Were treatment and control baskets similar before the campaign?
2. Observed composition: Which categories, brands, tiers and roles changed?
3. Commercial mechanism: Was the change addition, substitution, trade-up, promotion or trip-mix shift?
4. Causal estimate: Did the composition differ from a credible counterfactual?

IAB and IAB Europe distinguish attribution from incrementality and require credible counterfactuals and bias control.[2] This matters because media is often targeted using the same behavioral signals later used to describe “impact.”

Randomize treatment where feasible. If matching is needed, include pre-period basket structure, channel, store, mission proxies, promotion and availability. Do not compare exposed premium buyers with the total shopper base and call the resulting premium mix a campaign effect.

For the full event chain, use closed-loop measurement. Composition sits inside the purchase outcome, not in place of causal design.

Useful composition metrics

Choose a small set tied to the campaign job:

• target-product incidence;
• category share of units or value;
• distinct categories per basket;
• promoted-item share;
• private-label or brand share within category;
• premium-tier share;
• complementary-role attachment;
• substitute displacement;
• new category addition;
• margin contribution when available.

Report distributions or planned segments, not only overall means. One large mission group can hide opposite effects elsewhere.

Avoid creating a single “composition score” unless its components, weights and business meaning are explicit. A composite can simplify reporting, but it can also hide whether the change came from quantity, price, promotion or mix.

Common mistakes

Treating composition as intent. The basket suggests context but does not reveal the shopper’s thoughts.

Using unstable product taxonomy. Misclassified products create false category shifts.

Ignoring multi-mission baskets. One checkout can contain several unrelated tasks.

Confusing pair analysis with whole-basket structure. Attachment is one component, not the complete composition story.

Ignoring promotion and availability. Both can reshape the basket without media.

Inferring sensitive traits. Product patterns should not become speculative personal labels.

Reporting only value. Revenue can rise while category breadth, units or margin move differently.

Claiming incrementality from exposed baskets. Targeting often selects the composition it later “discovers.”

Buyer checklist

Before using composition in a plan or report, ask:

• Which dimensions define the profile?
• Are shares based on units, lines, revenue or margin?
• Is product taxonomy consistent across periods and channels?
• How are multi-mission baskets handled?
• Which products are complements, substitutes or unresolved?
• What campaign mechanism should change the mix?
• Are price, promotion and availability controlled?
• Are sensitive inferences prohibited?
• What comparison or holdout supports the impact claim?
• Which decision follows each composition pattern?

FAQ

Is basket composition the same as basket size?

No. Size measures quantity or breadth. Composition explains which categories, brands, tiers or roles make up the basket.

Is basket composition the same as cross-shopping?

No. Cross-shopping focuses on a defined relationship between entities. Composition describes the whole transaction structure.

Can basket composition identify a shopping mission?

It can provide evidence consistent with a mission, especially when combined with time, channel and sequence. It should not be treated as certain intent.

Should composition be measured by units or value?

Use the base that matches the decision, and often report both. Unit share and value share answer different questions.

What is the main retail media use?

To determine how the advertised product changed the basket: addition, substitution, trade-up, category expansion or no meaningful mix change.

Bottom line

Basket composition reveals structure, not certainty. Build the profile with clean taxonomy, separate quantity from value, account for multiple missions, and compare like with like. Then use a counterfactual to decide whether media changed the basket or merely selected a familiar one.

Sources

[1] Interactive Advertising Bureau and Media Rating Council, *IAB/MRC Retail Media Measurement Guidelines*, 2024.

[2] IAB and IAB Europe, *Guidelines for Incremental Measurement in Commerce Media*, 2025.

[3] Google Analytics, *Measure ecommerce*.

[4] arXiv, *Within-basket Recommendation via Neural Pattern Associator*, 2024.

[5] arXiv, *Inferring Complementary Products from Baskets and Browsing Sessions*, 2018.

[6] Allegro research authors, *Beyond Co-purchase Relation: Evolution of Complementary Recommendations at Allegro*, September 2026 preprint.

Ready to see how this works in practice?

Footprints AI helps brands and retailers measure what matters. See our customer success stories or get in touch to discuss your retail media strategy.