Comparison| AIpedia Editorial Team

AI Marketing Mix Modeling (MMM) Tools Comparison 2026: Recast, Mutinex, Measured and 7 More (Pricing & Free Plans)

We compared 10 AI marketing mix modeling (MMM) tools, including Recast, Mutinex, Measured and Haus, on pricing, free plans, Japanese support, platforms, strengths and weaknesses, with picks by goal and an FAQ (vendor websites checked in November 2026).

Choosing an AI marketing mix modeling (MMM) tool is hard because the tools differ in kind. Some are commercial platforms with advisors and dashboards. Others are open-source code that your own data team must run. Pricing also varies: a few vendors publish prices, most ask you to request a demo, and two tools are free to license. Those two still cost staff time and computing to run.

This article compares 10 tools: Recast, Mutinex, Measured, Haus, Keen Decision Systems, Paramark, Sellforte, Lifesight, Google Meridian, and Meta Robyn. For each one we cover pricing, free plans or trials, Japanese support, platforms, strengths, and weaknesses. We compared them using each vendor's official website, checked in November 2026.

The article also gives picks by goal and practical tips for buying without regrets. Check the official website for the latest pricing and features before you decide.

What Are AI Marketing Mix Modeling (MMM) Tools? What They Do and What to Know Before Choosing

AI marketing mix modeling (MMM) tools estimate how much each marketing channel contributes to sales or other results. They work from aggregated data, such as spend and outcomes over time, rather than individual-level tracking. Many tools now add experiments, forecasting, and AI agents that suggest actions. The results are statistical estimates, so treat them as one input to budget decisions, not a guaranteed answer.

Typical uses, based on the features listed on each tool's official website, include:

  • Measuring ROI by channel, including channels such as TV, out-of-home advertising, YouTube, and influencers (Paramark).
  • Running incrementality tests, such as geo experiments, and using the results to calibrate the model (Measured, Recast, Lifesight).
  • Optimizing budget allocation with scenario planning before you spend (Keen Decision Systems, Recast).
  • Forecasting sales and marketing targets under different scenarios (Recast).
  • Modeling promotions and seasonality, plus weather and events such as Black Friday, together with media (Sellforte).
  • Building your own model in-house with open-source code (Google Meridian, Meta Robyn).

Before choosing, check what each offer actually covers. Most commercial vendors do not publish prices and direct you to a demo or contact form. Keen Decision Systems offers a free trial, but its length and terms are not stated. Google Meridian and Meta Robyn are free open-source code, but they require staff who can work with Python or R, and computing to run the models.

Also check scale, platforms, and language. Haus and Measured position themselves for enterprises, while Sellforte lists monthly packages for retail and ecommerce brands. Paramark starts at $100k/year, which is a high bar for a small business. None of the 10 tools states a Japanese UI or availability in Japan on its official website, so confirm this directly with each vendor.

Comparison Table of Top 10 Tools (Pricing, Japanese Support, Platforms)

ToolMain plansJapanese supportPlatforms
RecastSee the official websiteNot statedWeb, API, MCP server
MutinexSee the official websiteNot statedWeb
MeasuredSee the official websiteNot statedWeb
HausBasics / Core / Plus / Enterprise: no prices listed (Get a demo); Causal MMM from PlusNot statedWeb
Keen Decision SystemsFree trial (length and terms not stated) / No prices listedNot statedWeb
ParamarkEssentials: $100k/year / Advanced: $150k/year / Enterprise: from $220k/year (billed annually)Not statedWeb, Slack, MCP
SellforteIncrementality Testing: $1,900/month / Full-Funnel MMM: $4,500/month + customization (example at $300,000 monthly media spend)Not statedWeb
LifesightContact salesNot statedWeb, MCP
Google MeridianFree (open source, Apache-2.0); staff and computing costs applyNot statedPython, GitHub (open source)
Meta RobynFree (open source, MIT); staff and computing costs applyNot statedR, Python API (beta), GitHub

AIpedia has not rated these tools, because they do not have their own pages on AIpedia yet. Pricing and plans change; we checked each vendor's official website in November 2026. Check the latest information on the official website before you sign up.

Features, Pricing, and Best For: 10 Tools

1. Recast

Recast is a platform for planning and analysis built on its own Bayesian MMM. According to the official website, it covers daily ROI by channel, promotion effects, GeoLift test design, and budget optimization. It models time-varying ROI, interactions between channels, promotion spikes, and multi-stage funnels.

  • Strengths: Tests and the model sit on one platform. GeoLift by Recast covers test planning, target-region selection, synthetic controls, and reading results, and feeds the results back into the MMM.
  • Strengths: The official website lists SOC 2 compliance and the Meta Measurement Partner badge. It also lists integrations with Slack, Notion, and Looker, plus an MCP server for Claude.
  • Strengths: Setup can be done directly with the Recast team or through a certified Operator.
  • Things to note: Pricing is not listed, and adoption assumes setup with the Recast team or an Operator, so it is hard to complete on your own.
  • Things to note: Figures such as forecast accuracy of 95% or more are the company's own claims with no third-party sources shown. The official website does not mention a Japanese UI or availability in Japan.

Pricing: Check the official website for pricing. The homepage does not list pricing and directs visitors to request a demo. It states that the pricing model may differ when going through an Operator (a consulting partner).

Best for: Teams that want to run geo tests and MMM together and are comfortable with vendor-assisted setup. It is a poor fit if you need public pricing or confirmed Japanese support.

2. Mutinex

Mutinex promotes an "agentic commercial mix model" that goes beyond MMM, supporting sales and marketing decisions. According to the official website, its main products are GrowthOS, MAITE, DataOS, and AI Signals. It states that it can deliver a production model from raw data within 24 hours.

  • Strengths: According to the official website, it is used by more than 100 global brands and supports more than $13B in commercial decisions. Customers listed include Optus, Samsung, Hershey's, ING, Asahi, and KIA.
  • Strengths: GrowthOS On-Demand is a self-serve version you can use on your own. DataOS automates data preparation for modeling.
  • Strengths: It publishes the Open MMM Validation Framework, an open-source framework for evaluating MMM providers.
  • Things to note: Pricing is not listed, and the official website names almost no specific integrations.
  • Things to note: Locations are Sydney, Melbourne, and New York, with no mention of a location in Japan or a Japanese UI, although Asahi is listed as a customer. Claims such as "up to 20% ROI improvement" are the company's own, and the conditions are not shown.

Pricing: Check the official website for pricing. The homepage does not list pricing. GrowthOS On-Demand, which you can use on your own, is listed.

Best for: Teams that want a self-serve option and a broader suite that includes an AI consultant and AI answer tracking. Confirm pricing, integrations, and language support with the vendor.

3. Measured

Measured is an enterprise AI marketing measurement platform that combines incrementality testing with MMM. According to the official website, it produces predictive ROI with a causal MMM calibrated by experiments. It emphasizes measuring all the way through to in-store sales.

  • Strengths: The causal MMM is calibrated by experiments, which aims to measure actual impact rather than correlation. Incrementality testing automates geo and audience split experiments.
  • Strengths: The official website lists more than 300 fully managed connections, plus SOC 2 Type II and ISO/IEC 27001:2022.
  • Strengths: Media Plan Optimizer lets you move from measurement to budget allocation. Cross-Channel Dashboard and Benchmarks support reporting and peer comparison.
  • Things to note: Pricing is not listed, and you request a demo. The official website positions it for enterprises and shows no sign that small-scale use is intended.
  • Things to note: Names of specific connections are not listed. There is no mention of a Japanese UI or availability in Japan. Case figures on the official website, such as Made In raising ROAS by 182%, are the company's own and shown without conditions.

Pricing: Check the official website for pricing. The homepage does not list pricing and directs visitors to request a demo. A readiness check (Check your readiness) is also offered.

Best for: Enterprise brands that want experiments and MMM in one platform, including retail brands that need to measure in-store sales.

4. Haus

Haus is an enterprise AI incrementality platform that measures the causal impact of marketing. According to the official website, it provides experiments, causal MMM, causal attribution, and Architect, a causal marketing agent. It states that it optimizes more than $30B in ad spend per year.

  • Strengths: Experiments, Causal MMM (updated weekly), and ad-level Causal Attribution are handled on one platform.
  • Strengths: Four plans let you choose by scale, from Basics for DTC on Shopify and Amazon to Enterprise with a dedicated data scientist.
  • Strengths: SOC 2 Type 2 and ISO 27001 badges are displayed. Customer examples include Grubhub, GrΓΌns, and FanDuel.
  • Things to note: No plan lists a price, and a demo request is required. Causal MMM is available only on Plus and higher plans.
  • Things to note: It targets enterprises, and Basics is described as best for self-starter DTC brands selling via Shopify and Amazon. There is no mention of a Japanese UI or availability in Japan.

Pricing: Contact sales. The official website describes four plans, Basics, Core, Plus, and Enterprise, but lists no prices. Each plan has a "Get a demo" prompt. Add-ons such as international support, additional experiment KPIs, retail breakouts, additional brands, additional Causal MMM models, and managed services have no pricing listed.

Best for: Companies that put incrementality experiments first and want MMM added at the Plus tier or above. DTC brands on Shopify and Amazon can look at Basics, which does not include Causal MMM.

5. Keen Decision Systems

Keen Decision Systems offers Keen OS, a marketing investment decision platform for deciding where to spend the next marketing dollar. According to the official website, more than 450 brands use it, and it has optimized $48B in marketing budgets. The headquarters is in Cary, North Carolina, in the US.

  • Strengths: Keen OS covers peer comparison (Intelligence), measurement (Measure), weekly scenario planning (Plan), and execution (Buy) in one flow.
  • Strengths: It optimizes from a profit perspective, using net profit and profit ROI.
  • Strengths: Integrations are listed specifically: The Trade Desk, Skai, Klaviyo, HubSpot Analytics, Snowflake, Meta Ads, LinkedIn Ads, TikTok, Criteo, and others. A free trial is offered.
  • Things to note: Pricing is not published, and the length and terms of the free trial are not stated on the official website.
  • Things to note: Figures on the official website are the company's own claims, and the conditions are not stated. There is no mention of a Japanese UI or availability in Japan.

Pricing: No prices are listed on the official pricing page. It offers a free trial, a demo request, and a contact form, but does not state the trial's length or terms.

Best for: Teams that want to try a platform before committing and care about profit-based planning and named integrations. Ask about the trial terms first.

6. Paramark

Paramark is a marketing measurement platform that combines incrementality testing, MMM, and forecasting. Human growth advisors and AI agents support planning, testing, and measurement. According to the official website, it measures ROI by channel with tests and MMM, including TV, out-of-home advertising, YouTube, and influencers.

  • Strengths: Prices are on the official pricing page, so you can compare plans up front. All plans include unlimited channels, tests, and users.
  • Strengths: According to the official website, model results become usable in a few weeks. You can work with human advisors on Slack.
  • Strengths: All plans include CSV downloads, and Advanced and above add MCP and warehouse sync. Open-source code is published on GitHub.
  • Things to note: Even the cheapest plan, Essentials, is $100k/year, which is expensive for small businesses. The listed prices are for annual billing.
  • Things to note: Essentials and Advanced cap the number of MMM and test metrics, and Growth Advisory is 1 to 3 hours per month depending on the plan. There is no mention of a Japanese UI or availability in Japan.

Pricing: Essentials: $100k/year (billed annually, 1 MMM metric, 2 test metrics, national model, 1 hour of Growth Advisory per month). Advanced: $150k/year (billed annually, 2 MMM metrics, 3 test metrics, hierarchical model, 2 hours of Growth Advisory per month, MCP and warehouse sync). Enterprise: from $220k/year (billed annually, custom quote). Add-ons: Brand Halo $50k/year, Data engineering onboarding $20k+/month. Annual billing is the cheapest, stated as up to 17% cheaper than billed monthly, quarterly, or semiannually.

Best for: Companies with a budget that fits annual contracts of $100k/year or more and that want advisor support.

7. Sellforte

Sellforte is a measurement and optimization platform for retail and ecommerce brands. According to the official website, it combines MMM, incrementality testing, and attribution into one always-on foundation. Brand customers listed include C&A, Lidl, Intersport, Tchibo, bonprix, Douglas, and Fressnapf.

  • Strengths: Monthly prices are listed, and you can choose tests, attribution, and MMM in stages.
  • Strengths: According to the official website, approved recommendations can be applied to Meta, Google, and TikTok via API. Changes are applied only after approval.
  • Strengths: Results can be exported directly to Snowflake and BigQuery. The MMM models promotions, seasonality, weather, and events such as Black Friday.
  • Things to note: Pricing varies with monthly media spend and sales channels, and $300,000 is only an example in the price table. Full-Funnel MMM and Enterprise have separate customization fees.
  • Things to note: The official website lists English as its only language and does not mention a Japanese UI. Multiple markets require an individual quote. A partnership with Exact AI is stated for APAC.

Pricing: Based on an example of average monthly media spend of $300,000: Incrementality Testing: $1,900/month. Incremental Attribution: $2,500/month. Full-Funnel MMM: $4,500/month + customization. Enterprise: $8,500/month + customization. Pricing is determined by the package, monthly media spend, and the sales channels covered. Prices can be shown in USD, EUR, GBP, and AUD. Contact sales for multiple markets.

Best for: Retail and ecommerce brands that want to start with a single package and add MMM later.

8. Lifesight

Lifesight is an integrated marketing measurement platform that combines causal MMM, continuous geo testing, and incrementality-adjusted attribution. AI agents make recommendations, and the team reviews them before acting. According to the official website, more than 300 brands use it.

  • Strengths: Geo tests are designed to calibrate MMM and attribution. Budget changes require team approval and can be automated if rules exist.
  • Strengths: The official website states SOC 2 Type II, ISO 27001, and GDPR compliance. Official partners are Google Ads, Meta, Amazon Ads, TikTok, and The Trade Desk.
  • Strengths: Lifesight MCP passes measurement context to Claude, ChatGPT, and others. Free tools such as the Measurement Gap Calculator are offered.
  • Things to note: Pricing is not published, and you get a proposal through a demo. MMM is included in the Precision plan and above.
  • Things to note: Results on the official website, such as a 5% increase in Walmart in-store sales, are the company's own case studies for individual customers. There is no mention of a Japanese UI or availability in Japan.

Pricing: Contact sales. No price is listed on the official pricing page. It is an annual subscription, quoted based on features, channel mix, number of brands and countries, data preparation, and scope of support. The three tiers are Performance, Precision, and Enterprise.

Best for: Teams that want MMM, geo tests, and attribution on one data foundation, with AI assistant connections through MCP.

9. Google Meridian

Google Meridian is an open-source MMM framework developed by Google. According to the official website, it is built on Bayesian causal inference, runs on aggregated data alone, and lets advertisers build and run models in-house.

  • Strengths: It is open source under Apache-2.0, with no license fee. It uses only aggregated data.
  • Strengths: Meridian GeoX designs geo experiments, and the results can be used as priors. Reach and frequency modeling connects video planning to results.
  • Strengths: The MMM Data Platform helps retrieve Google Ads and Display & Video 360 data, through an application form.
  • Things to note: You must set up a Python environment and build and run the model yourself. A GPU is recommended, and there is no official GPU support on macOS.
  • Things to note: It is not a commercial product with a point-and-click dashboard, so operation depends on your own data staff. No Japanese UI or availability in Japan is stated, though the documentation pages are displayed in multiple languages.

Pricing: Free (open source, Apache-2.0). No pricing is stated on the official website. Use of the MMM Data Platform is through an application form. Running the model still costs staff time and computing, and the official GitHub recommends at least one GPU.

Best for: Organizations with in-house staff who can work with Python and Bayesian modeling and want to avoid license fees. It supports Python 3.11 to 3.13.

10. Meta Robyn

Meta Robyn is an experimental, semi-automated open-source MMM package from Meta Marketing Science. According to the official website, it offers hyperparameter optimization with evolutionary algorithms, Ridge regression, decomposition with Prophet, and a budget allocation tool. It does not depend on personal information or cookies.

  • Strengths: It is open source under the MIT license, with no license fee. It is semi-automated and states the aim of reducing human bias in modeling.
  • Strengths: It can be calibrated with measured results such as geo tests and Facebook lift. It supports budget allocation and a one-page report for comparing models.
  • Strengths: The design does not depend on personal information, individual-level data, or cookie and pixel data.
  • Things to note: The official GitHub describes it as "experimental," and the Python API is in beta. It is mainly used with R, and setup requires R and the Python library Nevergrad.
  • Things to note: It is not a commercial product with a point-and-click dashboard, so operation depends on your own data staff. No Japanese UI or availability in Japan is stated.

Pricing: Free (open source, MIT). No pricing is stated on the official website. Running the model still costs staff time and computing.

Best for: Teams with R skills that want a license-free way to build MMM in-house and can accept an experimental package.

Recommendations by Goal

GoalRecommendedWhy
Use MMM with no license costGoogle Meridian, Meta RobynBoth are open source and free to license, under Apache-2.0 and MIT respectively. Staff and computing costs still apply.
Try a commercial platform firstKeen Decision SystemsThe official website offers a free trial, though its length and terms are not stated.
Compare published monthly pricesSellforteThe pricing page lists packages from $1,900/month to $8,500/month + customization, based on an example of $300,000 in monthly media spend.
Plan around published annual pricesParamarkEssentials is $100k/year, Advanced is $150k/year, and Enterprise is from $220k/year.
Enterprise incrementality testing with MMMMeasured, HausBoth position themselves for enterprises and combine experiments with causal MMM (on Haus, from the Plus plan).
Named integrations with existing systemsKeen Decision Systems, SellforteKeen lists tools such as Snowflake and HubSpot Analytics, and Sellforte exports to Snowflake and BigQuery.
Connecting to AI assistantsRecast, LifesightRecast has an MCP server for Claude, and Lifesight MCP supports Claude and ChatGPT.
A self-serve optionMutinexThe official website lists GrowthOS On-Demand, which you can use on your own.

None of the 10 tools states a Japanese UI or availability in Japan on its official website, so there is no pick for Japanese-language use. Confirm language support with each vendor before shortlisting.

How to Choose Without Regrets

Decide between commercial platform and open-source code

Google Meridian and Meta Robyn are free to license, but they run as code. Meridian needs Python, and Robyn is mainly used with R. Your own data staff must build, run, and maintain the models.

Commercial platforms such as Recast or Lifesight come with vendor support but have costs that are mostly not public. Compare the staff time you would spend with the fees you would pay.

Check how pricing is published

Paramark and Sellforte publish prices. Paramark lists annual plans from $100k/year, while Sellforte lists monthly packages from $1,900/month based on an example of $300,000 in monthly media spend.

Recast, Mutinex, and Measured do not list prices on their websites, and Haus, Keen Decision Systems, and Lifesight have pricing pages without amounts. Request quotes early, and ask what is included, such as MMM, advisory hours, or managed services.

Understand what the free trial and free open-source tools cover

Keen Decision Systems offers a free trial, but the official website does not state its length or terms. Ask before you rely on it for a full evaluation.

Open-source tools cost nothing to license. That does not include the cost of the people and computing needed to run them. Meridian recommends a GPU.

Match the tool to your scale

Measured and Haus position themselves for enterprises. Haus offers four plans, and Causal MMM is available only on Plus and higher plans.

Sellforte targets retail and ecommerce brands with staged packages. Paramark's lowest plan is $100k/year, so check it against your budget first.

Confirm language and platform fit

None of the 10 tools states a Japanese UI or availability in Japan on its official website. Sellforte's website lists English as its only language. Google Meridian and Meta Robyn are free to license, but for Japanese campaigns your team must prepare its own ad data and run the models in-house.

Also check the platform. Most commercial tools are web-based. Paramark adds Slack and MCP, Recast adds an API and an MCP server, and Meridian and Robyn run as code. Confirm the final decision with a qualified professional where legal or financial matters are involved.

FAQ

Q. Which AI marketing mix modeling tools in this comparison are free?

A. Google Meridian and Meta Robyn are open source and free to license. Google Meridian uses the Apache-2.0 license, and Meta Robyn uses the MIT license. Both are used as code rather than as commercial products, so running them depends on your own data staff and computing resources.

Q. Which tools publish prices?

A. Paramark lists Essentials at $100k/year, Advanced at $150k/year, and Enterprise from $220k/year. Sellforte lists Incrementality Testing at $1,900/month, Incremental Attribution at $2,500/month, Full-Funnel MMM at $4,500/month + customization, and Enterprise at $8,500/month + customization. Sellforte's prices are based on an example of average monthly media spend of $300,000.

Q. Which tools require a demo or contact to get a quote?

A. Recast, Measured, Haus, and Lifesight direct visitors to request a demo, and Haus lists four plans without prices. Keen Decision Systems offers a contact form and a free trial, but the length and terms of the trial are not stated. Mutinex does not list pricing on its official website, although it lists GrowthOS On-Demand as a self-serve version. Check the official website for the latest pricing and features.

Q. Do these tools support Japanese?

A. For Recast, Mutinex, Measured, Haus, Keen Decision Systems, Paramark, Lifesight, and Meta Robyn, the official websites do not state a Japanese UI or availability in Japan. Sellforte's website lists English as its only language. Google Meridian does not state a Japanese UI either, though its documentation pages are displayed in multiple languages. Mutinex lists Asahi, a Japanese beverage company, as a customer.

Q. Which tools combine experiments with MMM?

A. Recast, Measured, Haus, Paramark, Sellforte, and Lifesight all describe testing alongside MMM. Recast offers GeoLift by Recast, and Measured calibrates its causal MMM with experiments. Haus includes Causal MMM only in Plus and higher plans. Google Meridian and Meta Robyn can also be calibrated with experiment results such as geo tests.

Q. What should I check before choosing a tool?

A. Check your budget scale first, since Paramark's cheapest plan is $100,000 per year. Then check whether you have data staff for Google Meridian (Python) or Meta Robyn (R). Many figures on the official websites, such as accuracy and ROI gains, are company claims without stated conditions. Confirm the final decision with a qualified professional where legal or financial questions apply.

Summary

The 10 tools fall into three groups. Paramark and Sellforte publish prices. Recast, Mutinex, Measured, Haus, Keen Decision Systems, and Lifesight do not list prices, and most direct you to a demo. Google Meridian and Meta Robyn are free open-source code that needs in-house data staff and computing.

Scale and language matter as well. Measured and Haus position themselves for enterprises, and Paramark starts at $100k/year. None of the tools states a Japanese UI or availability in Japan, so confirm this before buying.

  • Want to avoid license fees β†’ Google Meridian (free open source; needs Python skills and computing)
  • Want to try a platform first β†’ Keen Decision Systems (free trial offered, terms not stated)
  • Want to compare published prices β†’ Sellforte (monthly packages listed on the pricing page)

View all AI marketing tools (14 items)

πŸ“ˆBrowse all tools in AI Marketing β†’

Comparisons for Other Categories

Is your tool missing from this article? Vendors can add it with a paid section written by our editors (JPY 50,000, one-time), or take the first position, labeled β€œSponsored” (JPY 80,000). If a fact about a tool here is wrong, we correct it for free.

Written & verified by

A

AIpedia Editorial Team

The AIpedia Editorial Team specializes in researching, comparing, and hands-on testing AI tools. We create accounts and use the tools we cover, verifying pricing, key features, and real-world usability before writing. Articles are reviewed regularly to keep the information up to date.