Digital Lexicon Hub

A statistical method that estimates how each marketing channel and external factor contributes to sales, using aggregated historical data.

MMM measures channels that user-level attribution cannot track, including TV, offline media and walled gardens, without relying on cookies, so it guides annual budget allocation and shows where extra spend stops paying back.

A retailer feeds Meridian three years of weekly sales by region, spend for eight channels, prices, promotions and holidays, calibrates paid social with a geo lift test and moves 15% of budget from saturated search to video.

Results depend on data volume and spend variation, so channels with steady spend or short history get uncertain estimates. Models show associations shaped by assumptions, and calibration with Incrementality or Conversion Lift experiments is needed for reliable decisions.

MMM regresses a KPI such as sales on media inputs and controls such as price, seasonality and macro factors, usually with weekly or geo-level data. Adstock transforms capture carryover effects and saturation curves capture diminishing returns. Open-source options include Google's Meridian, which supports reach and frequency inputs and calibration with geo experiments, and Meta's Robyn, which uses ridge regression with Nevergrad optimization. MMM complements an Attribution Model rather than replacing it.