{"page":"/blog/retail-media-measurement-standards-explained","modified":"2026-10-18","license":"https://creativecommons.org/licenses/by/4.0/","columns":["Method","How it works","Strengths","Limitations","Named in"],"rows":[["Randomised holdout (RCT)","Randomly split the eligible audience into exposed and held-out groups; compare outcomes. The exposed group is an intention-to-treat group, since not every member will see the ad (IAB/MRC, ch. 4)","Least biased baseline; the reference method","Needs scale, a holdout the sales team will accept, and clean separation of exposed and unexposed; cannot measure off-platform halo","IAB/MRC ch. 4; IAB/IAB Europe Nov 2025 (experiments); IAB in-store (test vs control at store level)"],["Geo test","Treat and withhold media by geography (stores, regions) and compare","Works where user-level holdouts are impossible, including in-store and offsite","Regions differ; needs pre-period matching and controls for price, promotion and feature space (IAB in-store, Dec 2024, test-vs-control best practices)","IAB in-store Dec 2024; IAB Oct 2024 (match-market testing)"],["Matched-market test","Pair test and control markets on pre-period sales, size, format and range; analyze with difference-in-differences","Practical for networks with store estates; transparent","Imperfect matches bias the estimate; must exclude stores with other activity (IAB/MRC, ch. 4)","IAB/MRC ch. 4; IAB Oct 2024"],["Synthetic control","Build a counterfactual for the exposed group from the purchase behavior of unexposed consumers with similar pre-period characteristics, often with propensity-score matching","No holdout needed; usable after the fact","Depends on observed covariates; the guidelines stress that a synthetic control is not synthetic data (IAB/MRC, ch. 4)","IAB/MRC ch. 4; IAB/IAB Europe Nov 2025 (model-based counterfactuals)"],["Marketing mix modeling (MMM)","Econometric model of sales against media, price, promotion and seasonality over time","Cross-channel view; the US guidelines ask networks to supply data for brands' MMM (IAB/MRC, §3.5)","Low granularity; slow; needs long history","IAB/IAB Europe Nov 2025 (econometric models); IAB Oct 2024"],["Machine learning counterfactuals","Models trained on deterministic exposure, outcome and covariate data to estimate what would have happened","Scales to many campaigns","Opaque unless documented; MRC transparency on probabilistic data applies (IAB/MRC, ch. 4)","IAB/MRC ch. 4"],["Hybrid proxies","Combine a small experiment with modelled extrapolation","Cheaper than full experiments","Must disclose the extrapolation rules","IAB/IAB Europe Nov 2025"],["Before-and-after (sales variance)","Compare sales before, during and after","Simple","Not incrementality; the in-store standard allows it only when incremental studies are not possible (IAB in-store, Dec 2024, §6)","IAB in-store Dec 2024 (as a fallback)"],["Shadow-mode testing","Run the decisioning without serving, to measure what the model would have changed","Useful for algorithm changes","Not an outcome measurement on its own","IAB Oct 2024 (IAB, 31 October 2024)"]],"sourceUrls":[["https://www.iab.com/wp-content/uploads/2024/01/IAB_Retail_Media_Measurement_Guidelines_January2024.pdf"],["https://www.iab.com/wp-content/uploads/2024/12/IAB-Europe_IAB-In-Store-Retail-Media-Standards_FINAL-DECEMBER-2024.pdf"],["https://www.iab.com/wp-content/uploads/2024/01/IAB_Retail_Media_Measurement_Guidelines_January2024.pdf"],["https://www.iab.com/wp-content/uploads/2024/01/IAB_Retail_Media_Measurement_Guidelines_January2024.pdf"],["https://www.iab.com/wp-content/uploads/2024/01/IAB_Retail_Media_Measurement_Guidelines_January2024.pdf"],["https://www.iab.com/wp-content/uploads/2024/01/IAB_Retail_Media_Measurement_Guidelines_January2024.pdf"],[],["https://www.iab.com/wp-content/uploads/2024/12/IAB-Europe_IAB-In-Store-Retail-Media-Standards_FINAL-DECEMBER-2024.pdf"],["https://www.iab.com/guidelines/retail-media-advanced-measurement-and-data-collaboration/"]]}