PromptSharp › Prompt Library › CPG › Incrementality test design for retail media — before you spend the budget

Incrementality test design for retail media — before you spend the budget

The network reports a strong ROAS and you cannot tell how much of it you would have earned anyway.

The prompt — copy and run it

You are a commerce media analyst designing an incrementality test for a retail media investment. I will paste the current plan. Produce:

A) A TEST DESIGN — the holdout method available on this platform (geo, audience, or store-level), the split, the minimum duration, and what specifically is withheld from the control.
B) A MEASUREMENT PLAN — the primary metric, the guardrail metrics, the sample condition under which the read is valid, and the explicit statement of what result would be indistinguishable from noise.
C) A CONTAMINATION LIST — everything that could leak between test and control (shared shoppers, national promo, organic search halo, other media) and the mitigation for each, with any that cannot be mitigated named as a caveat on the final read.

Inputs: [PASTE PLATFORM, BUDGET, CAMPAIGN TYPES, GEOGRAPHY, CURRENT REPORTED ROAS, ANY PRIOR TEST RESULTS]

Rules: Do not invent, estimate, or extrapolate any figure — if a number is not in the data I give you, write "not provided" and flag it. Mark every claim I should verify against my syndicated data or internal reporting before using it externally. Never include retailer-confidential terms or personally identifiable shopper data. Never treat platform-attributed ROAS as incrementality; state the difference explicitly in the output.

Why this prompt works

Retail media reporting is graded by the party selling the media. Designing the holdout — and naming in advance what a null result looks like — is the only way the number that reaches the CFO means what the deck says it means.

Frequently asked

Related CPG prompts