Marketing mix modeling (MMM)
Marketing mix modeling is a way to guess how much each ads channel added to sales, using weeks of spend and sales numbers, without following individual people. It helps you plan next quarter, such as more YouTube or more Instagram. It is not a tool for changing today's bids. Lemonado is not a marketing mix modeling product.
Theodor Lindfors, Founding Marketer ·
Why MMM is back
People also search this as media mix modeling. Same tool: a model of weekly spend and sales, not a person-level credit split. Cookies and device IDs got worse. Person-level attribution (who gets credit for this order) lost signal. MMM (marketing mix modeling) never needed a person ID. Open-source tools made it cheaper. Teams now use it for questions like: more YouTube or more Meta (Facebook and Instagram) next quarter?
Let's take a drink brand as an example. They sell a low-sugar sports drink for active women over 50, mostly from their own website. The drink brand buys Meta (Facebook and Instagram), Google, TikTok, and YouTube. An MMM (marketing mix modeling) takes two years of weekly spend by channel, weekly website sales, and (if they are honest) promo weeks and price changes. It guesses how much of each week's extra cases came from each channel, how long an ad keeps working after you stop paying, and how returns shrink as you spend more on the same place.
MMM (marketing mix modeling) is a planning tool. Let's take a drink brand as an example. They sell a low-sugar sports drink for active women over 50, mostly from their own website. An MMM will not tell the drink brand which Instagram headline to run on Tuesday. It will tell them whether last quarter's mix was buying extra cases or buying credit. That is a different job from attribution inside an ads manager. Lemonado is not an MMM product. It will not run this model for you.
How to read an MMM
MMM (marketing mix modeling) is a model of history. It will not tell you which keyword to add. Pair it with incrementality tests (holdouts that measure extra sales) so the model is not only fitting last year's budget process. Let's take a drink brand as an example. They sell a low-sugar sports drink for active women over 50, mostly from their own website. If the drink brand has always spent the most on Meta (Facebook and Instagram), a naive model will keep saying Meta did the work, because that is where the money was.
You need clean weekly spend and sales, by channel, for a long enough history to run an MMM (marketing mix modeling). Memorial Day promos, a price cut, and New Year fitness season belong in the model or they will look like media. Garbage in still produces a confident chart. Twelve messy weeks is not two years.
Common MMM mistakes
- Treating MMM (marketing mix modeling) ROAS (return on ad spend) as a bid target inside Google Ads. Let's take a drink brand as an example. They sell a low-sugar sports drink for active women over 50, mostly from their own website. The drink brand should not paste a quarterly model into Tuesday's bids.
- Fitting a model on 12 messy weeks and calling it proof.
- Leaving promos out, then deciding YouTube does not work because every YouTube spike sat on a sale week.
For the weekly question (why did Meta CPA, or cost per acquisition, jump?), use analytics. For the quarterly question (more YouTube or more Meta?), MMM (marketing mix modeling) is in the conversation. Keep MER (marketing efficiency ratio: all drink revenue divided by all ad spend) next to it so a confident model cannot outrun company revenue. Lemonado feeds the weekly job live. It does not replace the quarterly model. Lemonado is not an MMM product.
Lemonado
How Lemonado helps with marketing mix modeling
Lemonado is not an MMM (marketing mix modeling) product. An MMM still needs clean weekly spend and revenue by channel. Lemonado already sits on those sources, so the inputs live in one place instead of a quarterly CSV hunt. For day-to-day questions (why did Meta CPA, or cost per acquisition, jump?), use analytics.