Retail media return on investment, with attribution labelled as allocation

Retail media return on investment needs a causal contribution estimate and a complete cost boundary. A report assigning sales to adverts is not enough. For an England campaign on retailer or marketplace controlled inventory, separate transaction records, attribution allocation, causal evaluation and future projections before calculating a return.

The return calculation below draws on desk research completed on 6 September 2026. No campaign outcomes, costs or incremental-sales estimates were collected. Every numerical input remains UNKNOWN, and no positive return, saving or benchmark is implied. Finance and evaluation specialists must review any later calculation.

Fix the commercial perspective

State whether the decision concerns the advertiser's campaign or the retailer's advertising operation. Advertiser product contribution and retailer advertising revenue are not interchangeable benefits. Choose a defined activity, period and set of costs, then preserve that boundary through the calculation.

Identify what happens without the activity. A comparison against no campaign can answer a different question from a comparison against another placement. The counterfactual should match the decision being evaluated rather than change when a preferred result appears difficult to support.

Keep recorded outcomes in their own account

Record what the transaction system actually reports, including its date range and treatment of cancellations, returns and tax. Preserve the source file and reconciliation status. A recorded purchase is an observation; it does not state whether advertising caused the purchase.

Do the same for delivery events. Ad requests, served impressions, viewable impressions and clicks describe different quantities. None becomes a sale because it appears next to a revenue column, and no unknown quantity should be inferred from an unsupported average conversion rate.

Label attribution as allocation

Specify the attribution rule, lookback period, eligible touchpoints and sales definition. Keep attributed sales in a separately labelled result. Changing the allocation method can change that result without demonstrating a corresponding change in underlying shopper behaviour.

The IAB Europe May 2026 standards address attribution and incrementality as distinct measurement topics. This industry framework does not supply an England campaign's causal result. A ratio of attributed sales to advertising spend must retain that description rather than be relabelled profit ROI.

Establish what a causal estimate can support

The Magenta Book analytical-methods annex describes experimental and quasi-experimental evaluation approaches. Its UK government context supplies methodological guidance, not a legal requirement for private retail-media campaigns.

Ask the evaluator to choose a design that fits the proposed question and available evidence. Document allocation or comparison construction, contamination risks and uncertainty. A before-and-after change alone should not be treated as the causal answer when other explanations have not been addressed.

This article reports no experiment, statistical test or matched comparison. If the evidence cannot support incremental sales, label that result UNKNOWN. Do not manufacture an uplift merely to connect an attribution dashboard to a budget spreadsheet.

Calculate only a defined contribution return

Use this explicit modelling convention: G is incremental contribution before campaign costs, and C is incremental campaign cost for the identical scope and period. The net-return ratio equals (G minus C) divided by C. It is a proposed accounting definition, not a measured result or universal supplier metric.

G must reflect the agreed product costs and returns without already deducting the campaign costs in C. Have finance confirm VAT treatment and avoid counting the same expense twice. Where either value is unsupported, or C is not positive, the ratio remains UNKNOWN.

Keep projections apart from evaluation

A forward scenario may use documented assumptions, but label its output projected. Retain the source and owner of every assumed input, including any margin or future quantity. Do not present a projected return as recorded performance or a causal estimate from a completed study.

Present the decision with its evidence limitations and unanswered questions. This is not financial or tax advice, and no commercial outcome is guaranteed. Obtain named reviewer approval and recheck the methodological sources before publishing a return claim.