How to Read an MMM Report Without a Statistics Degree

You just got an MMM report. It is 20 pages of charts, tables, and terms like "posterior distribution" and "adstock decay." Your eyes glaze over. Here is what actually matters and what you can safely ignore.

Channel Contribution: Where Did Revenue Come From?

This is the headline number. Channel contribution shows how much of your total revenue (or conversions) each marketing channel drove. It also shows your "baseline" — the revenue that would have happened without any marketing at all (from brand awareness, organic traffic, repeat customers, etc.).

A typical breakdown might look like: Baseline 55%, Google Ads 18%, Meta 12%, TV 8%, Email 4%, Other 3%. If your baseline is above 60%, your marketing is a smaller driver than you might think. If it is below 40%, your marketing engine is doing heavy lifting.

ROAS by Channel: How Efficient Is Each Dollar?

ROAS (Return on Ad Spend) tells you how many dollars of revenue you get for each dollar spent. A ROAS of 3.0 means $1 in spend generates $3 in revenue.

But watch out: ROAS in an MMM report is different from ROAS in Google Ads or Meta Ads Manager. Platform-reported ROAS includes double-counted conversions and non-incremental revenue. MMM ROAS measures the incremental impact only. It is almost always lower than what platforms report, and that is the accurate number.

If your MMM shows Meta ROAS at 1.8x but Meta Ads Manager says 4.2x, the truth is closer to 1.8x. Plan your budget accordingly.

Response Curves: When to Stop Spending

Response curves (also called saturation curves) are the most actionable chart in any MMM report. They show the relationship between spend and incremental revenue for each channel.

The curve starts steep (each additional dollar generates a lot of return) and gradually flattens (each additional dollar generates less). The point where the curve starts to flatten is where diminishing returns kick in. Anything past that point is inefficient spend.

Example: if your Google Ads response curve flattens around $150K/month, spending $200K means the last $50K is generating significantly less return than the first $150K. That $50K might do better on a channel that has not hit saturation yet.

Confidence Intervals: How Much to Trust the Numbers

In Bayesian MMM, every number comes with a range. When the report says "Meta ROAS: 1.8x (1.2x - 2.5x)," the 1.2x-2.5x part is the 90% credible interval. It means the model is 90% confident the true ROAS falls somewhere in that range.

Narrow intervals mean the model is confident. Wide intervals mean there is not enough data to be sure. If one channel shows ROAS of 3.0x (0.5x - 6.2x), that range is so wide it is almost meaningless. You need more data or a different approach for that channel.

Rule of thumb: if the lower bound of the interval is below 1.0x, there is a real possibility that channel is losing money.

Adstock and Carryover: The Lag Effect

Not all marketing impact happens immediately. TV ads create awareness that leads to conversions weeks later. This lagged effect is called "adstock." The report will show an adstock decay rate for each channel, usually as a half-life.

A half-life of 2 weeks means half the impact of an ad occurs in the first 2 weeks, with the rest tapering off over the following weeks. Digital channels typically have a half-life of 1-3 days. TV and brand campaigns can be 2-4 weeks.

Why this matters: if you pause TV ads and see no immediate revenue drop, that does not mean TV was not working. The effect takes weeks to fade. Do not make the mistake of cutting a channel based on same-week performance alone.

Want to see all of these concepts in a real report? Explore our interactive demo.