Marketing Performance Management Software
Robyn: Open-Source Media Modeling
facebookexperimental.github.io · checked 30 Sept 2026
Open-source media modeling for evaluating channel efficiency, setting budgets, and projecting results.
How it plugs in
Signal on Marketing Performance Management Software6.3#2 of 24
Meta Marketing Science offers Robyn as an open-source marketing mix modeling package. It applies machine learning and time-series methods to estimate channel effects, compare models, calibrate results with experiments, and reallocate budgets. It is free, privacy-friendly, and suited to digital or direct-response advertisers with detailed data. Stable R, beta Python, API, and self-hosted options are available.

Review
In a line
Robyn is an experimental open-source marketing mix modeling package. It helps estimate how media channels affect results and supports budget decisions.
What it does
Using machine-learning methods, Robyn measures media-channel efficiency and effectiveness, along with adstock rates and saturation curves. Adstock describes how advertising effects carry over time; saturation shows when extra spend produces smaller gains.
It can model trend, seasonality, and holidays, then compare models with one-page summaries. Forecasting, scenario planning, ROI reporting, and budget allocation are included.
A constrained optimization solver reallocates budgets toward stronger projected outcomes. Robyn can also calibrate models against geo-based tests, Facebook Lift, MTA, randomized controlled trials, and geo experiments.
How it connects
Robyn is MIT-licensed open-source software. The maintained R release is published on CRAN, while development work appears on GitHub; a beta Python rewrite is also documented.
It uses Nevergrad for hyperparameter optimization, Prophet for time-series decomposition, and glmnet for ridge-regression fitting. Running the beta Python interface first requires installing the R package.
Robyn runs through an API and supports Linux, macOS, Windows, and self-hosted setups. The maker describes it as not requiring personally identifiable information, individual-level log data, cookies, or pixel data. Paid-media variables and spend vectors must be equal in length and follow the same media sequence.
Good fit
Robyn is aimed at digital and direct-response advertisers working with rich, granular datasets containing many independent variables. It is useful for measuring media returns, testing scenarios, forecasting, and planning budgets without individual-level tracking.
Watch for
Robyn is experimental, so teams should check whether its workflow fits their data and modeling process. The Python edition is an LLM-translated beta and may contain bugs or translation problems.
The Python API also depends on the R package. Media inputs and their spend vectors must align in both length and order. Support is provided through GitHub issues and a public Robyn MMM Users Facebook Group.
Pricing
Robyn is free. Its free plan has no listed charge, and the package is MIT licensed.
Verdict
Choose Robyn if you need an open-source way to analyze marketing channels, run forecasts, compare models, or plan budgets from detailed aggregate data. Its privacy design avoids personally identifiable and individual-level data.
Look elsewhere if you need a non-experimental product, a mature Python-only workflow, or inputs that cannot meet Robyn’s strict media-vector ordering. Python users must install R first, and the beta may include bugs.
Our take
Robyn fits data-rich marketing teams seeking free, open-source media measurement and budget planning. Its R release is stable, while the Python option is still beta and tied to the R package. The package avoids cookies, pixels, and personal data, which can simplify privacy reviews. However, it is experimental and expects carefully structured media inputs. Organizations seeking a fully mature Python product or a turnkey commercial service should consider other marketing measurement tools instead.
- Free and MIT-licensed open-source package
- CRAN hosts a stable R release
- Supports forecasting, scenarios, ROI reporting, and budget allocation
- Uses aggregate data without PII, cookies, or pixel data
- Can calibrate results with experiments and lift methods
- Experimental project with a beta Python rewrite
- Python API depends on the R package
- Media variables and spend vectors must align in length and order
- Built for granular, data-rich datasets
- Beta Python version may contain bugs
- Pick it ifDigital and direct-response advertisers working with rich data sources; Teams needing open-source media modeling and budget planning; Analysts comfortable with R or a mixed R/Python setup
- Skip it ifYou need a mature Python-only product without an R dependency; Your media inputs cannot follow Robyn’s required ordering and lengths; You want a turnkey commercial service rather than an experimental package
Pricing
Robyn is free, with no listed subscription charge. Its MIT license also permits open-source use under that license.
Questions
Is Robyn free?
Yes. Robyn is offered as a free plan and is MIT-licensed open-source software.
Which languages can I use with Robyn?
Robyn has a stable R version on CRAN, an R development version on GitHub, and a beta Python rewrite. The Python interface requires the R package first.
What data does Robyn need?
It is built for granular datasets with many independent variables. Paid-media variables and paid-media spend vectors must be equal in length and use the same media order.
Does Robyn use personal data or cookies?
No. Robyn is designed to work without personally identifiable information or individual-level log data, and it does not depend on cookies or pixel data.
What marketing tasks does Robyn support?
It supports media-effect modeling, forecasting, scenario planning, ROI reporting, model comparison, calibration, and budget allocation.
How does Robyn handle seasonality?
Facebook Prophet automatically separates trend, seasonal effects, and holidays in its time-series processing.
Where can I get help with Robyn?
The maker points users to GitHub issues and a public Robyn MMM Users Facebook Group.
Near it
All alternatives →- GMGoogle Meridian#1 on this board
- PMPyMC-Marketing#3 on this board
- AC
Adobe Customer Journey Analytics#4 on this board - RMRevup Marketing#5 on this board
- KKeen#6 on this board
- C
Cassandra#7 on this board - SSellforte#8 on this board
- AAprimo#9 on this board
- B
Budgee#10 on this board - EBEtropo Budget Buddy#11 on this board