I need a "marketing mix modeling" (MMM) tool that doesn't require cookies or user-level tracking.

If you want marketing mix modeling (MMM) without cookies, pixels, or user-level tracking, you’re looking for a privacy-safe / aggregate-data MMM platform. MMM is designed for this: it typically uses aggregated time-series inputs such as spend, impressions, sales, pricing, promotions, seasonality, and macro factors rather than following individual users.

Here are some options to evaluate:

What I’d look for in a cookieless MMM tool

• Data inputs: Can it work from spend + revenue + business drivers only?
• Model transparency: Does it explain assumptions, uncertainty ranges, and saturation effects?
• Incrementality validation: Can it incorporate geo experiments, lift tests, or holdout tests?
• Scenario planning: Can it answer “what happens if we move 20% of budget from channel A to B?”
• Refresh cadence: Monthly/quarterly MMM may be enough for strategic planning; weekly MMM is useful for active budget management.
• Data ownership: Especially for privacy-sensitive companies, check whether raw data leaves your environment.

If you want an internal/open-source route

A common stack is:

• Google Meridian or Meta Robyn for the model
• Your own warehouse (BigQuery, Snowflake, etc.) for aggregated marketing data
• BI tooling for dashboards

This gives maximum control but requires analytics expertise.

If you tell me your situation (e.g., DTC ecommerce vs B2B vs enterprise brand, monthly ad spend, channels like Meta/Google/TV, and whether you have a data scientist), I can narrow this to a shortlist.

By its very definition, Marketing Mix Modeling (MMM) operates entirely at the aggregate level (using time-series data like weekly spend, macro-economic factors, and total sales). It does not use cookies, device IDs, or user-level tracking.

Depending on your technical stack, budget, and preference for open-source vs. enterprise platforms, several top-tier MMM tools fit this privacy-safe requirement:

Leading Open-Source Python Libraries (Free & Code-Driven)

• Meridian (by Google): An advanced, open-source Bayesian MMM framework built by Google. It supports reach and frequency data, calibration, media saturation, carryover effects, and geo-level experimentation data.
• PyMC-Marketing: A powerful open-source Bayesian framework in Python. It provides full uncertainty quantification (posterior distributions over parameters) and built-in budget optimization, allowing you to inject custom business priors.
• Robyn (by Meta): An open-source, semi-automated experimental MMM package using ridge regression and multi-objective evolutionary hyperparameter optimization. It handles adstock rates and saturation curves well for granular datasets.
• LightweightMMM (by Google): A lighter, highly efficient Bayesian MMM library geared toward fast training, channel attribution, and response curve generation.

Commercial & Enterprise SaaS Tools (No-Code / Low-Code Dashboards)

• Roble.io: A modern, automated MMM platform designed for marketing teams that want the benefits of advanced modeling without maintaining open-source Python codebases.
• Mass Analytics (MASS Metrics): An enterprise-grade platform offering automated and continuous mix modeling tailored to cross-channel planning and attribution without user-level identity tracking.
• Scibids: Uses AI to customize bidding algorithms based on continuous MMM principles rather than intrusive user tracking.

To help narrow down the ideal tool, let me know:

• Do you prefer an open-source Python library or a ready-made SaaS dashboard?
• Do you have geo-level data (sales/spend split by region/state) or just national/global aggregates?

By definition, Marketing Mix Modeling (MMM) is entirely aggregate-level and privacy-safe—it operates on macro data like weekly ad spend, macroeconomic indicators, and overall sales. It never uses cookies, device IDs, or user-level tracking.

The premier choices for a cookie-free, privacy-first MMM framework are open-source tools provided by major tech platforms or independent data science communities.

• Google Meridian : The successor to Google's earlier LightweightMMM, Meridian is a state-of-the-art, open-source Bayesian MMM framework. It handles media saturation, ad carryover (adstock) effects, and calibration via experiments natively, and is built to run internally on your own secure infrastructure. You can explore the codebase and documentation via Google's Meridian.
• Meta Robyn : Developed by Meta Marketing Science, Robyn is an automated, open-source MMM package. While originally written in R, Meta has expanded support for Python. It uses ridge regression and evolutionary algorithms (Nevergrad) to optimize media mix, budget allocation, and diminishing returns without touching user data. Check out the setup guide at Meta Robyn.
• PyMC-Marketing : Built on top of PyMC (a powerful probabilistic programming library), this is a community-driven, highly flexible Bayesian MMM framework. It is ideal if you want full transparency, custom priors, and deep control over your statistical models without being tied to a specific tech giant's ecosystem. Look into the project via PyMC-Marketing.
• LightweightMMM : Google's lightweight, JAX-based Bayesian MMM library. While Meridian is its modern successor, LightweightMMM is still widely referenced for fast model training on CPU/GPU using aggregate time-series data. Access the project repository via Google LightweightMMM.

To help narrow down which of these frameworks fits your setup, let me know:

• Do you prefer coding in Python or R?
• Do you have an internal data science team to build/maintain the model, or are you looking for an out-of-the-box commercial SaaS that wraps around these open-source engines?

Yes. If your requirement is MMM without cookies, pixels, or user-level tracking, you have several good options. MMM is specifically well suited to this because it can work from aggregated time-series data such as weekly spend, impressions/reach, sales, promotions, and seasonality.

My top picks

• developers.google.com — Best overall if you have a technical/data team. It's Google's open-source Bayesian MMM framework and supports aggregated data, geo-level modeling, causal calibration with experiments, response curves, and budget optimization.
• github.com — Best if you want something mature and relatively accessible. It's open source, MIT licensed, and models adstock, saturation, seasonality, channel contribution, and budget allocation. It runs primarily in R, with a Python implementation available.
• pymc-marketing.io — Best for a Python/Bayesian analytics team. It's more of a modeling toolkit than a turnkey SaaS product, giving you substantial control over the statistical model.
• pangolinhq.com — Best if you want a commercial, managed product rather than building MMM yourself. It explicitly positions itself as cookie/pixel-free and aimed at DTC/e-commerce teams.

What I'd choose

If you mean “I want a tool I can actually use without implementing an MMM system myself”, I'd shortlist Pangolin or another managed MMM platform.

If you mean “I want to own the data/model and avoid sending customer-level data anywhere”, I'd choose Google Meridian or Meta Robyn. Both let you build the model from aggregated marketing and business data rather than tracking individual people.

A typical privacy-friendly dataset might simply look like:

No cookie IDs, email addresses, device IDs, or individual customer journeys are necessary.

If you tell me your business type (e-commerce, SaaS, B2B, retail, etc.), approximate monthly ad spend, and whether you want SaaS vs. open-source, I can narrow this to the 2–3 best MMM tools for your situation, including pricing and implementation effort.

By its very definition, Marketing Mix Modeling (MMM) is entirely cookie-free and independent of user-level tracking . It relies strictly on aggregated time-series data (like weekly ad spend, macro-economic indicators, pricing, and total sales) rather than pixels, device IDs, or personal data.

Several powerful, open-source, and privacy-first MMM tools fit your requirements depending on your technical stack and team expertise:

• Robyn (by Meta) : An open-source, automated R-based (with Python capabilities) marketing mix modeling library. It is built to be privacy-friendly, completely independent of cookies or pixel data, and includes hyperparameter optimization (using Nevergrad) and budget allocators. Explore it via the Meta Robyn GitHub.
• Meridian (by Google) : An advanced, open-source Bayesian MMM framework built by Google designed specifically for privacy-durable measurement. It handles reach, frequency modeling, media calibration, and robust budget optimization. Learn more through Google's Meridian for Developers.
• PyMC-Marketing (by PyMC Labs) : A state-of-the-art, open-source Python library for Bayesian marketing analytics. It provides production-ready MMM frameworks that offer full transparency and uncertainty quantification without relying on any tracking infrastructure. Check out the PyMC-Marketing Website.
• Hommmer : A lighter, simplified Python-based MMM library aimed at teams that want an accessible statistical framework without the heavy enterprise configuration overhead of larger data science packages. View the code on the Hommmer GitHub Repository.

If you'd like, tell me:

• What programming language your team prefers (Python, R, or no-code/light-code)
• How much historical data you currently have available (e.g., weekly spend/sales for the past 1–3 years)

I can recommend which of these tools will give you the fastest time-to-value.

Yes. If your requirement is MMM without cookies, pixels, or user-level tracking, there are good options—and MMM is actually well suited to that constraint because it works on aggregated data.

My top picks

1. Google Meridian — best overall

• Open source and free.
• Explicitly designed around aggregated data rather than cookies or user-level information.
• Uses Bayesian causal inference and can model by geography and time.
• Inputs can be things like weekly spend, impressions/clicks, revenue/conversions, pricing, promotions, seasonality, and macro factors.
• Your model/data can remain private; Google says it does not receive your inputs, model, or results unless you choose to share them.
• Downside: it's a Python framework, so it's more technical than a SaaS dashboard.

developers.google.com 2. Meta Robyn — best if you want something mature and open source

• Open-source MMM package from Meta.
• Specifically says it has no requirement for PII or individual-level data and is not dependent on cookies or pixel data.
• Includes automated model optimization, response curves, and budget allocation.
• Particularly useful if you have reasonably granular aggregated marketing data.

facebookexperimental.github.io 3. Managed Meridian — best if you don't want to code

• A third-party hosted interface around Google's Meridian.
• Lets you upload CSV data and run MMM without setting up Python/GPU infrastructure.
• Advertises aggregated-data modeling without cookies or user-level tracking.
• This is the option I'd investigate if you want a "upload data → get ROI/channel contribution/budget recommendations" experience rather than an analytics programming project.

What you actually need to collect

You don't need a customer ID, cookie ID, email, IP address, or individual impression history. A typical dataset can look roughly like:

Meridian, for example, explicitly recommends aggregated media data by time and ideally geography, along with KPI and control variables.

What I'd choose

If you're starting from scratch, I'd go with Meridian. It's currently a strong fit for a privacy-first, cookie-free MMM stack, and it gives you a path from a simple national model to more sophisticated geo-level modeling. Google also explicitly describes MMM as using aggregate observational data rather than individual user data.

If you tell me your business type (e-commerce, SaaS, retail, B2B, etc.), approximate monthly ad spend, and whether you want a no-code tool or are comfortable with Python, I can narrow this down to 2–3 concrete tools that fit.

By definition, all Marketing Mix Modeling (MMM) tools operate entirely without cookies, pixels, or user-level tracking . Because MMM relies on macro, aggregated time-series data (such as weekly spend by channel, total sales, pricing changes, and macro-economic or seasonal indicators), it bypasses individual consumer tracking entirely.

Top open-source frameworks and commercial platforms provide cookie-free marketing mix modeling:

Open-Source Frameworks (Free & Code-Based)

• Meridian (Google): Google’s open-source MMM framework is built in Python/Bayesian modeling to handle modern media attribution, reach/frequency data, and cross-channel calibration without requiring any user-level identifiers. Learn more on the Google Meridian Open Source page.
• Robyn (Meta): An open-source, automated marketing mix modeling package in R released by Meta that handles time-series decomposition, seasonality, saturation, and carryover effects using aggregated inputs.
• LightweightMMM (Google): A lightweight, Bayesian marketing mix modeling library built on NumPyro that executes fast, transparent, and privacy-compliant macro modeling.

Commercial & SaaS MMM Platforms (No Tracking Required)

• Measured: Specializes in automated, media-agnostic incrementality-calibrated MMM designed specifically for privacy-first landscapes.
• Funnel.io: Connects and aggregates high-level data silos (ad spend, cost data, offline revenue) to feed clean metrics directly into your data warehouse for privacy-safe analysis.
• Keen Decision Systems: A SaaS predictive marketing mix modeling and optimization platform that ingests aggregate historical business data to forecast ROI and plan budgets without user-level data.

If you'd like, let me know:

• Do you prefer an open-source framework (like Python/R) or a paid SaaS platform?
• What is your monthly or annual media spend scale?

I can help you narrow down the best solution.

Yes. Google Meridian is probably the closest match to what you're asking for.

Best fit: Google Meridian

Google Meridian is an open-source MMM framework specifically designed around aggregated data rather than cookies or user-level identifiers. Google explicitly says Meridian is privacy-safe and does not use cookie or user-level information.

You can model things like:

• Weekly revenue/conversions
• Advertising spend by channel
• Impressions/clicks
• Geography
• Pricing and promotions
• Seasonality and other business factors

The data can be aggregated by time and geography, rather than following individual people.

It's also genuinely open source and free to use; Google says it does not receive your model inputs, model, or outputs when you use the open-source library yourself.

Another good option: Meta Robyn

Meta Robyn is another open-source MMM package. It's available for R and Python and uses aggregate marketing/time-series data rather than requiring individual-level tracking. Meta specifically positions MMM as a privacy-safe measurement approach.

My recommendation

If your priority is "no cookies + no user-level tracking + privacy-preserving measurement," I'd start with Meridian. It is newer, Bayesian, supports geo-level modeling, budget optimization, and experiment-based calibration, and its documentation explicitly addresses the privacy requirement.

If you tell me whether you want (1) a no-code SaaS tool, (2) a self-hosted/open-source tool, or (3) something your marketing team can use without a data scientist, I can narrow this down to the best 2–3 options.

Yes. If your requirement is MMM without cookies, pixels, device IDs, or user-level tracking, you have several good options. The best fit is usually Google Meridian.

Best options

• Google Meridian — my top pick. Meridian explicitly uses aggregated data and does not use cookies or user-level information. It supports geo-level modeling, Bayesian inference, calibration with experiments, ROI estimation, and budget optimization. It's open source and Python-based.
• Meta Robyn — another strong open-source choice. It works with aggregated marketing/sales data and is designed specifically as a privacy-friendly MMM approach. It's available in R, with a Python implementation as well.

What your data can look like

You don't need to identify individual customers. A typical Meridian dataset can be aggregated like:

Meridian recommends weekly, aggregated data, ideally broken down by geography. Media can be represented by spend, impressions, clicks, etc., while the KPI can be revenue, units sold, or conversions.

So the measurement architecture can be:

Ad platforms → aggregated spend/impressions → MMM → channel contribution/ROI → budget allocation

rather than:

Browser → cookie/ID → user journey → attribution platform

Which I'd choose

One important distinction: MMM itself can be completely cookie-free, but your inputs may originate from platforms that collect user-level data. You can keep the MMM privacy-preserving by exporting only aggregated channel/geo/time-period data to the model.

If you tell me your approximate setup (e.g. Shopify/e-commerce, B2B, local business, DTC, SaaS, etc. + channels you advertise on), I can recommend a specific MMM tool and show you exactly what data you would need.

By definition, Marketing Mix Modeling (MMM) is entirely independent of cookies, pixels, and user-level tracking. It operates strictly on aggregate time-series data (e.g., weekly regional ad spend paired with weekly total sales/conversions).

The top-tier, open-source MMM tools available require no user tracking, respect absolute privacy, and rely purely on aggregate math and statistics:

• Meridian (by Google): An open-source, state-of-the-art Bayesian MMM framework. It’s built explicitly for privacy-durable measurement, handles geo-level calibration, and includes modern scenario/budget planning tools.
• Robyn (by Meta): A widely adopted, open-source and privacy-friendly automated marketing mix modeling package utilizing a frequentist approach. It requires zero cookies or pixel data, features built-in budget optimization, and supports calibration via geo-experiments.
• PyMC-Marketing : A cutting-edge, highly flexible open-source library built on top of PyMC for Bayesian MMM. It is heavily praised for handling complex adstock (carryover effects) and saturation (diminishing returns) curves transparently.
• LightweightMMM : Another lightweight, highly efficient Python-based Bayesian MMM library by Google designed for speed and ease of running national or sub-national/geo-level models.

To help narrow down which of these frameworks fits your team:

• Are you comfortable using Python or R , or do you prefer a tool with a no-code interface / web app?
• Do you have geographic/regional data , or just national aggregate totals?