Marketing Mix Model (MMM) Calibration

Use incrementality evidence to calibrate your MMM and inform channel iROAS estimates.

Geo tests

Conversion lift studies

A/B tests

Priors

Bell curve illustrating a prior distribution
MMM

Sellforte calibrates MMM with incrementality evidence. Here's how it works

Bayesian MMM starts with priors

Sellforte uses Bayesian Marketing Mix Modeling. Priors describe the plausible range of iROAS for each channel before the model learns from your data.

Prior distributionA bell curve shows a range of plausible channel iROAS values before the model is fitted to the data. Probability density Channel iROAS
Prior distribution

Calibration makes priors informative

Instead of using uninformative priors, Sellforte creates informative priors from incrementality evidence. The evidence shapes both the expected iROAS and the uncertainty around it.

Uninformative prior and calibrated priorA broad uninformative prior is compared with a narrower calibrated prior informed by incrementality evidence. The distributions are illustrative. Probability density Channel iROAS
Calibrated priorUninformative prior

The model estimates final iROAS

MMM combines the calibrated prior with your marketing and sales data to estimate iROAS. The resulting posterior distribution captures the model’s estimate and its uncertainty.

Calibrated prior and posterior from modelingThe calibrated prior is shown with a posterior distribution that reflects the prior and the data used by the model. The distributions are illustrative. Probability density Channel iROAS
Calibrated priorPosterior from modeling

We prioritize incrementality evidence based on its relevance to the activity being modeled

  1. Priority 1

    Channel-specific incrementality tests

    Geo tests, conversion lift tests, and A/B tests specific to the channel that is being calibrated.

  2. Priority 2

    Incrementality evidence from comparable markets

    Incrementality tests from comparable markets for the channel being calibrated.

  3. Priority 3

    External benchmarks

    Incrementality evidence from similar activities in comparable industries.

Full self-serve calibration solution for the marketing analytics and data science teams

Analyze incrementality tests

Review results from geo tests, Conversion Lift Studies, and A/B tests. Assess incremental ROI, uncertainty, and test scope before using the evidence to calibrate your MMM.

Explore incrementality testing →

iROAS and uncertaintyOriginal iROAS probability curve from the Sellforte demo. Mean iROAS 5.01, with 90 percent interval 4.21 to 5.81. The shaded area marks that interval. CONVERSION LIFTiROAS and uncertaintyMeta Cart RetargetingTest iROAS5.0190% interval4.21–5.8100.40.844.555.56

Add all the tests to your test library

Add your geo tests, Conversion Lift Studies, and A/B tests to one shared library. Keep test results and their channel and market context together, ready to use in MMM calibration.

Explore the Test Library →

Test libraryYour incrementality test results
TestSpendiROASConfidence
Meta Cart RetargetingConversion Lift · United States$61.11k5.0196%
Meta + Google go-darkGeo test · United States$1.08M5.06100%
Email winback pushA/B Test · United States$7.63k9.50100%
Meta full-funnelConversion Lift · United States$350.41k3.1498%
Meta ProspectingConversion Lift · United States$360.61k2.6595%

Connect tests to MMM media features

Link each relevant test to the media feature it informs. Review the connected experiments and their ROI estimates together, so the calibration evidence matches the channel being modeled.

Media featuresConnect tests to media features
Media featureExperiment
iROAS		90% interval
Meta Advantage+ · RevenueMeta Ads · US · Short uplift
4.263.90–4.61
Connected experimentsiROAS90% interval
Meta US upper-funnel test4.213.79–4.63
Meta Advantage+ US test4.403.74–5.06

Configure calibration parameters by advertising channel

Configure the remaining calibration parameters for each advertising channel. Review evidence sources, prior assumptions, weights, and other settings in one table.

Channel settingsCalibration parameters by channel
Advertising channel
Performance MaxGoogle AdsValue not shownValue not shownValue not shownValue not shownValue not shownValue not shownValue not shownValue not shownValue not shown
ShoppingGoogle AdsValue not shownValue not shownValue not shownValue not shownValue not shownValue not shownValue not shownValue not shownValue not shown
Search GenericGoogle AdsValue not shownValue not shownValue not shownValue not shownValue not shownValue not shownValue not shownValue not shownValue not shown
Search BrandGoogle AdsValue not shownValue not shownValue not shownValue not shownValue not shownValue not shownValue not shownValue not shownValue not shown
ProspectingMeta AdsValue not shownValue not shownValue not shownValue not shownValue not shownValue not shownValue not shownValue not shownValue not shown
Advantage+Meta AdsValue not shownValue not shownValue not shownValue not shownValue not shownValue not shownValue not shownValue not shownValue not shown

Review prior iROAS and MMM iROAS

Review prior iROAS and MMM iROAS side by side for each advertising channel. Compare the calibrated prior with the model’s final estimate, including uncertainty, to see how your marketing and sales data updated the starting assumptions.

Prior vs. MMMCompare prior and MMM iROASMeta Sales · Meta Ads · US · Revenue
Prior iROAS
3.20
MMM iROAS
3.48
Calibrated prior2.00–4.40
MMM estimate2.70–4.26
iROAS

Prioritize the evidence your MMM needs next

Use Sellforte’s Test Prioritizer to identify gaps in incrementality test coverage and rank testing opportunities across channels and markets.

Review the Impact, Confidence, and Ease behind each recommendation. Combine the ranking with your campaign calendar and testing capacity to build a roadmap for gathering new calibration evidence.

As tests are completed, review the results and use relevant evidence to refine your MMM calibration.

Explore the Test Prioritizer

Test prioritizerRanked testing opportunities
ImpactConfidenceEase
#Media & marketWhy it ranksPriority / 10
1Meta SalesMeta Ads · Germany9.6
2Meta SalesMeta Ads · United States9.2
3Meta ProspectingMeta Ads · Germany9.2
4Meta Advantage+Meta Ads · United States9.1
5Meta ProspectingMeta Ads · United States9.1

Calibration available in Sellforte MCP

Calibration evidence
You

Which tests inform our Meta ROI?

AI

Let’s review the studies used for calibration.

Calibrated priors
You

What is the prior for this channel?

AI

Here’s the calibrated ROI range and its uncertainty.

Prior vs. posterior
You

How did modeling update the prior?

AI

Let’s compare the prior with the final ROI distribution.

Market benchmarks
You

Can we use another market’s tests?

AI

Let’s compare the channel, market, and study scope.

Evidence gaps
You

Which channels need more testing?

AI

Let’s find gaps in relevant incrementality evidence.

New test results
You

How does the new test compare?

AI

Let’s compare it with the existing calibration evidence.

Calibration report
You

Explain our calibration to the team.

AI

Here’s a summary of the evidence, priors, and model ROI.

Business context
You

How does our campaign plan affect this?

AI

Let’s review calibration alongside your campaign context.

Review your calibration evidence

Use your AI tools to explore the incrementality studies behind your model. Review the channels and markets covered by the evidence, the ROI estimates, and their uncertainty.

Compare priors and model results

Bring calibrated priors and MMM estimates into the same conversation. Explore how the final ROI compares with the evidence used to inform it, with channel and market context.

Create calibration reports

Turn calibration evidence into presentations, documents, and clear explanations for your team. Keep the contributing studies, priors, and model estimates together through Sellforte MCP.

Plan your next incrementality tests

Identify channels where evidence is limited or less relevant. Combine calibration data with your campaign plans to prioritize the next geo test, Conversion Lift Study, or A/B test.

Explore Sellforte MCP

Model Calibration FAQs

What is MMM calibration?

Calibration uses relevant external information, such as incrementality experiments, to inform model assumptions or parameters. In Bayesian MMM, this can take the form of priors about channel effectiveness.

Which studies are retrieved automatically?

The automated workflow described here uses the native Meta connection to retrieve Meta Conversion Lift studies. Other experimental evidence can inform the broader calibration process through the appropriate setup and review.

Can geo tests inform the model?

Yes. Relevant geo tests can inform MMM priors. The outcome, media intervention, geography, and period need to be aligned with the model. The automated Meta workflow should not be assumed to apply identically to every test type.

Can attribution data also inform calibration?

Sellforte’s calibration framework can use attribution data as additional information. Attribution and randomized experiments provide different kinds of evidence, which need to be treated appropriately in the model design.

Will a single new study overwrite the model?

The evidence is reviewed and weighted within the calibration workflow. Bayesian estimation combines prior information with historical data; the final model estimate is not a direct replacement with one study’s point estimate.

What if a study disagrees with the MMM?

First align the scope and definitions, then review the study design, uncertainty, data, and modeling assumptions. A disagreement can identify a question for investigation or further testing.

Get a deep-dive to MMM calibration with a Sellforte expert

Explore how incrementality evidence informs your MMM and discuss calibration options for your channels and markets.

Contact us

Subscribe to the Sellforte Impact Newsletter