Robyn
Robyn is an open-source Marketing Mix Modeling package from Meta Marketing Science for estimating how media channels perform. It uses machine-learning techniques to model channel efficiency, effectiveness, adstock rates and saturation curves. Robyn automates hyperparameter optimization with evolutionary algorithms and uses ridge regression to address multicollinearity and reduce overfitting. Its time-series modeling uses Prophet to account for trend, seasonality and holidays. Models can be calibrated against methods including geo experiments, Facebook Lift and MTA. A budget allocator uses a constrained nonlinear solver to propose reallocations aimed at maximizing outcomes, and model one-pagers help compare outputs. Robyn is designed for granular datasets with many independent variables, especially those used by digital and direct-response advertisers with rich data. It does not require personally identifiable or individual-level data and does not depend on cookies or pixel data. A stable R version is available on CRAN, with a development version on GitHub. The Python version is a beta rewrite that may have translation issues; its API requires the Robyn R package to be installed first.
Who it is for
Robyn suits digital and direct-response advertisers working with granular datasets and many independent variables. It is intended for teams that can work with its R package or evaluate a beta Python version.
What is good
- Estimates channel performance, adstock and saturation.
- Automates hyperparameter optimization with evolutionary algorithms.
- Can calibrate models against experiments and other methods.
- Budget allocation supports constrained reallocation scenarios.
- Does not require individual-level data, cookies or pixel data.
What to know first
- The Python version is beta and may have translation issues.
- The beta Python API requires the Robyn R package installed.
- Paid media variables and spend vectors must match in length and order.
Verdict
Robyn offers a modeling and budget-planning workflow for advertisers with detailed marketing data, alongside privacy-conscious input requirements. The stable R version is distinct from its beta Python option, which has stated limitations.
Compared on marketing performance management software
- Free plan
- Yes
- Budget planning
- Yes
- Forecasting
- Yes
- Scenario planning
- Yes
- ROI reporting
- Yes
