Marketing Mix Modeling (MMM)

Marketing Mix Modeling (MMM) is a statistical method that estimates how much each marketing channel contributes to sales, using data aggregated by week: spend per channel, revenue, price, promotions and seasonality. The results are used to decide how to split next quarter’s budget across channels.

What is marketing mix modeling?

MMM answers questions at the level of the whole business. If we cut YouTube ads this month, how much in sales do we lose? The model lines up sales (or leads, or app installs) against spend per channel over 2 to 3 years, separates out the sales that came from Tet (Lunar New Year), discount campaigns and competitor moves, then assigns what’s left to each channel.

Because MMM only uses aggregated numbers, it needs no cookies or device identifiers, and it can account for TV and out-of-home advertising too.

How does MMM work?

• Carryover effect (adstock): this week’s ads keep having an effect for several weeks afterward.
• Saturation: doubling spend doesn’t double sales. Past a certain threshold, extra money produces almost no extra orders.
• Control variables: seasonality, price and branded search volume, so the model doesn’t wrongly credit organic sales to advertising.

The outputs include ROI per channel, response curves and budget allocation scenarios. The two most widely used open-source libraries are Google’s Meridian and Meta’s Robyn. Meridian uses a Bayesian model and accepts geo-level data as well as reach and frequency data for video (Google, Introduction to Meridian, July 2026). Google opened Meridian to all marketers on January 29, 2025 (Google Ads & Commerce Blog, January 2025). Robyn uses ridge regression, handles seasonality with the Prophet library and comes with a built-in budget allocator (Meta Robyn, accessed October 2026).

How much data do you need?

The Meridian documentation gives an example. A national-level model with 12 channels and 6 control variables, run on 2 years of weekly data (104 data points), has only about 4 data points per parameter, too few to trust. Merge down to 3 channels, drop unnecessary variables and use 3 years (156 weeks), and that rises to about 15 data points per parameter, enough for directional results (Google, Amount of data needed, June 2026).

Why is MMM getting so much attention in 2025 and 2026?

Tracking individuals keeps getting harder, especially on iOS since ATT, and Meridian and Robyn are free. The industry still isn’t satisfied. In an IAB survey of more than 400 brand and agency leaders, 60% to 75% of those using advanced measurement methods said the methods fall short on rigor, speed and reliability, and none believed today’s MMM fully captures every paid channel. IAB has launched Project Eidos, which includes building common standards for MMM (IAB, State of Data 2026, February 2026).

How is MMM different from incrementality testing?

Incrementality testing runs an experiment with a control group for one channel over a set period. MMM looks at every channel at once, but it relies on observational data and can go wrong when several channels rise and fall together. Experiment results can be fed into MMM as starting values (priors) to calibrate it.

Example from Vietnam (hypothetical)

A household goods brand sells through its website, Shopee and physical stores, and has 3 years of weekly data. Inside Google Ads, Search has the highest ROAS. MMM shows Search is close to saturation, while YouTube and TikTok video drive extra branded searches one to two weeks later. The plan: keep Search as is, put the additional budget into video, and test it first in a few provinces.

Frequently asked questions

Can a small business do MMM?

Yes, if it has 2 to 3 years of weekly data and spend that varies enough. If the budget stays flat all year, a few incrementality tests will give a clearer answer.

How often should MMM be rerun?

Usually every quarter or every six months, in line with the budgeting cycle. Daily bids are still adjusted using platform data.

If you want to know whether the data you already have is enough to build an MMM, Adtek’s Lead Generation team can review it with you via our contact page.