Marketing mix modelling (MMM) is a statistical method that estimates how much each channel of advertising really added to a brand's sales. It uses months or years of total sales and total spending, not data about individual people. It is also called media mix modelling. Software products now run it for brands every week, where it once took a team of analysts several months.
The model starts from a simple idea. Sales go up and down every week for many reasons: prices, festivals, the weather, competitors and advertising. If the brand records all of these for a long enough period, statistics can separate them. The model estimates the sales the brand would have made with no advertising at all. These are called base sales. It then estimates how much each channel added on top. This extra is called the incremental effect.
Two patterns make the model realistic. An ad keeps working for a few weeks after it runs, because people remember it. This is called carryover, or adstock. And each extra rupee in one channel adds less than the one before, because the same people see the ad again. This is called saturation.
Consider a packaged tea brand that sells across eastern India. It feeds the model three years of weekly data. The data includes its sales from retail measurement, its spending on TV, YouTube, search ads and shop displays, its prices, and the dates of Durga Puja and the monsoon. The model finds that TV added about 9% of sales, and that YouTube added 6%. But spending on YouTube beyond ₹40 lakh a month added almost nothing more.
The brand checks the result with an experiment. It stops its YouTube ads for four weeks in six cities, and keeps them in six similar cities. Sales in the cities without ads fall by 5%, close to what the model predicted. Experiments of this kind are called geo experiments, because they compare places.
Why is MMM back in use? Attribution follows one person from an ad to a purchase. Data protection laws and phone makers now block much of that tracking. MMM follows no one, so it still works. It also measures TV and posters, which attribution cannot track at all.
MMM has a weakness. It needs a lot of history, and it gives estimates with a margin of error, not exact answers. A mobile measurement partner gives fast, detailed answers for each ad, but cannot say whether the sale would have happened anyway. Many brands use both. The spending figures usually come from marketing reporting and analytics.