Media mix modelling is a statistical approach that, based on historical data, estimates how individual marketing channels contribute to sales. It works with aggregated data – weekly channel spend, revenue, price, seasonality, competitor promotions and other influences – and does not need to track individual users, which is its main advantage in an environment with limited measurement. It can also include offline channels such as radio, television or outdoor advertising, and capture the delayed effect of brand campaigns. Its disadvantage is data intensity: you need a consistent history, generally at least two years, and sufficient variability in spend for the model to have something to compare. For smaller companies, a geo experiment tends to be more practical, giving an answer for a single campaign faster and more cheaply.
See also: Incrementality, Geo experiment, Company marketing budget.
