-1.png?width=1132&height=292&name=Asset%201@2x%20(12)-1.png "Odins.ai Logo")
Glossary
Marketing mix modeling, term by term.
Thirty-three terms you meet when you work with marketing mix modeling, from adstock to marginal ROAS. Each one with a plain definition, how it works, and how it shows up in a budget decision.
Reviewed November 2026
The model
How the model is built.
What a Bayesian marketing mix model is made of, and what each part means for how much the data can teach it.
Marketing mix modeling (MMM)
The statistical method behind everything else on this list: spend in, sales out, every channel on equal terms.
Read the definitionBayesian marketing mix modeling
A model that starts from what you already know and lets the data correct it, with a range on every answer.
Read the definitionPrior
What the model believes about a channel before it sees the data. The part of MMM that decides how much the data can teach.
Read the definitionPosterior
The model's answer: the prior after the data has had its say, with a most likely value and a range.
Read the definitionConfidence range
The range a Bayesian model reports around every number. A recommendation without one is hiding something.
Read the definitionControl variable
The non-marketing factors the model explains sales with, so they are not credited to a channel. Fewer than you would think.
Read the definitionBaseline (base sales)
The sales that would have happened without paid marketing. Everything the model credits to a channel is measured above it.
Read the definitionChannel response
How channels respond.
The shapes the model estimates for every channel: where returns flatten, when effects land, and how the year changes both.
Saturation curve
The response curve per channel: how sales respond to spend, and where the next unit stops paying.
Read the definitionDiminishing returns
Each extra unit of spend in a channel earns a little less than the one before. Why the average is the wrong number to plan on.
Read the definitionAdstock
The carryover effect: the part of this week's advertising that lands in the weeks that follow.
Read the definitionSeasonality
The yearly pattern in demand that the model separates from marketing, so a December spike is not credited to December's ads.
Read the definitionTime-varying saturation
Letting a channel's response curve follow the year, so November and July are not forced to share one curve.
Read the definitionLong-term brand effect
The slow lift in baseline demand from sustained brand investment, beyond the weeks adstock covers.
Read the definitionReturns and budgets
Reading the results.
The numbers a budget decision is made on, and why the average is usually the wrong one.
ROAS
Return on ad spend: revenue per unit of spend. Useful, as long as you know which ROAS you are looking at.
Read the definitionMarginal ROAS
What the next unit of spend in a channel earns at today's level. The number a budget change actually asks for.
Read the definitionMarginal CAC
The cost of the next customer in a channel at today's spend. The CAC version of marginal ROAS, for businesses that count customers.
Read the definitionIncremental sales
The sales a channel caused that would not have happened without it. The only kind of sales worth paying for.
Read the definitionChannel contribution
What each channel added to sales above the baseline over a period. The decomposition behind every ROAS figure.
Read the definitionBudget optimization
Finding the split and the total where the next unit works hardest, under the constraints the business has.
Read the definitionScenario planning
Forecasting what a budget would return before committing it, and comparing several futures side by side.
Read the definitionForecast verification
Forecast first, compare with actuals afterwards. The only honest way to know whether a model is right.
Read the definitionMeasurement methods
MMM next to attribution and experiments.
What each method measures, where it breaks, and how the three combine inside one model.
Attribution
Crediting conversions to the touchpoints that preceded them. Great inside a platform, not a currency across them.
Read the definitionLast-click attribution
All the credit to the final click. Simple, free, and systematically wrong about which channels create demand.
Read the definitionIncrementality test
A controlled experiment that measures what one channel causes, by comparing exposed and unexposed groups or periods.
Read the definitionGeo experiment
Run the activity in some regions and not others, then compare. The classic causal test, when the market is big enough to split.
Read the definitionPulse test
Vary one channel's spend over time, on and off or up and down, so the model can see how it responds.
Read the definitionTriangulation
Combining MMM, attribution and experiments. The question is where: inside one model, or in a dashboard afterwards.
Read the definitionCalibration
Anchoring a model to outside evidence, usually experiment results. In a Bayesian model, calibration is a prior.
Read the definitionData and operations
Getting the data in, and the answers out.
The structure every source lands in, how offline spend is collected, and how the model is kept current and used.
Channel hierarchy
The one structure every source lands in: market, product, segment, then channel down to the spot. Digital and offline side by side.
Read the definitionOffline media spend
TV, radio, out-of-home, print, influencer. No API, so it needs a pipeline, down to the individual spot.
Read the definitionStructured marketing data
Every source, one dataset, your hierarchy, your warehouse. The foundation the model sits on, and a product on its own.
Read the definitionModel retrain
Data flows in daily; once a month the model is trained again on everything so far. Slow, stable movement is the point.
Read the definitionMCP (Model Context Protocol)
The open standard that lets your own AI assistant talk to Odins: run scenarios, pull data, read the model, from Claude, ChatGPT or Copilot.
Read the definitionFrequently asked questions.
What is marketing mix modeling, in one sentence?
A statistical method that measures how marketing spend drives sales from aggregated data, with no user-level tracking, and separates each channel's effect from everything else that moves demand. The full definition, or the complete guide.
What is the difference between MMM and attribution?
Attribution credits the touchpoints before a conversion from tracked data; it is the best tool for optimizing inside a platform. MMM measures what each channel caused above the baseline from aggregated spend and sales, offline included; it is the source for total budget and channel split. Attribution explains where each applies.
What is the difference between ROAS and marginal ROAS?
ROAS is what a channel earned per unit of spend on average over a period. Marginal ROAS is what the next unit would earn at today's spend, read off the saturation curve, and it is the figure budget decisions belong on. Marginal ROAS.
Why are modern marketing mix models Bayesian?
Because marketing data is thin: often one data point a week to estimate roughly ten channels. A Bayesian model starts from what is already known about each channel, as priors, and lets the data update it, so it stays stable on little data and reports a range on every answer. Bayesian marketing mix modeling.
Where should I start if I am new to MMM?
Read marketing mix modeling, then saturation curve and marginal ROAS. Those three carry most budget conversations. For the whole picture in one place, the complete guide to marketing mix modeling.
See these terms on your own numbers.
A 30-minute walkthrough of saturation curves, marginal returns and scenarios, on a model built for channels like yours.
Book a demo