Robyn

Robyn is an open-source marketing mix modeling package from Meta Marketing Science for teams analyzing advertising performance across many variables. It is best suited to digital and direct-response advertisers with granular data and the expertise to work with a technical modeling package. Its stable R release is the sensible starting point; the Python option is still beta.

Overview

Robyn combines media modeling, experiment-based calibration and budget allocation in a free package. It estimates channel efficiency and effectiveness, adstock and saturation, then helps compare models and plan budget shifts. That breadth makes it useful for measurement teams building their own workflow, but it is not a ready-made, low-effort solution for advertisers with limited data or modeling capacity.

Robyn is designed for granular datasets with many independent variables, especially those from digital and direct-response advertising. Its privacy-oriented design does not require personally identifiable information or individual-level log data, and does not rely on cookies or pixel data. Paid media variables and spend vectors must have matching lengths and media order, so careful input preparation matters.

Key features

Robyn estimates media effects with machine-learning techniques and offers geometric, Weibull CDF and Weibull PDF adstock transformations to represent how advertising impact persists. Nevergrad evolutionary algorithms automate hyperparameter optimization, while ridge regression helps regularize multicollinearity and prevent overfitting. These methods address common modeling challenges in complex datasets, but they do not make input quality or model interpretation irrelevant.

For time-series structure, Robyn uses Prophet to decompose trend, seasonality and holiday patterns, with glmnet used for ridge regression fitting. It can calibrate models against ground-truth approaches including geo-based tests, Facebook Lift and MTA, and supports causal calibration such as randomized controlled and geo experiments. This gives teams a way to anchor model estimates in observed outcomes, provided they have suitable experiments.

The budget allocator uses a gradient-based constrained nonlinear solver to maximize outcomes by reallocating budgets. Model one-pagers support intuitive comparisons, while budget planning, forecasting and scenario planning support downstream decisions. These capabilities make Robyn relevant beyond estimation, though allocation recommendations still depend on the underlying model and its inputs.

Pricing

Robyn is free, with a free plan and no paid plan described. That removes a software subscription cost for teams able to operate the package themselves; it does not change the need for appropriate data, technical capacity or valid measurement inputs.

Platforms

Robyn supports API, Linux, macOS, Windows and self-hosted use. The R package has a stable CRAN version as well as a development version on GitHub. The Python version is a beta translation of the R package and may have bugs or translation issues; its API also requires the Robyn R package to be installed first. For production work, the R release is the more mature choice, while Python users should be prepared for those dependencies and limitations.

Who it's for

Robyn is a strong fit for digital and direct-response advertisers with rich, granular media data, analysts who need to compare marketing mix models, and teams that can calibrate estimates with experiments. Its privacy design is useful where individual-level records, cookies or pixel data are not wanted. It is a weaker fit for teams seeking a turnkey service, minimal setup, or a mature Python-only workflow.

Pros and cons

• Pros: Free and open source under the MIT license, allowing teams to use the package without a software subscription.
• Pros: Combines adstock options, automated optimization, calibration and constrained budget allocation in one modeling workflow.
• Pros: Does not require PII, individual-level logs, cookies or pixel data.
• Cons: Its focus on granular, high-variable datasets and aligned media inputs places a real data-preparation burden on users.
• Cons: The Python version is beta, may encounter bugs, and depends on the R package being installed.

Alternatives

For another free, open-source marketing mix modeling option, Google Meridian offers source code and methodology papers intended to support transparency and user auditability. Compare it with Robyn if auditability of the implementation and methodology is a priority.

Revup Marketing has a free plan limited to one plan, one user and one team, with higher tiers for more plans, users and teams; consider it if those plan-based limits better suit your workflow. Etropo Budget Buddy is a paid budget-planning alternative, with an Essential plan at $89 a month; choose it if you want a budget-planning product rather than Robyn's open-source modeling package.

Keen is another paid alternative. LiftLab offers an enterprise marketing measurement platform with custom pricing and terms specified in an Order Form, making it a consideration for teams seeking an enterprise platform. Planful uses custom pricing tailored to company size, user count and selected modules. Revlo Marketing is another paid alternative. Cassandra is a paid option with a six-month minimum commitment; its Essentials plan costs €2,700 a month and includes up to three MMM models and five incrementality experiments per year, so it is a substantially different commitment from free Robyn.

Browse the Marketing Performance Management Software category for more options.

Verdict

Choose Robyn if your team has granular media data, can work with a technical modeling package and wants free, open-source modeling with calibration and budget allocation. Its combination of modeling tools and privacy-conscious inputs is the main reason to choose it. Look elsewhere if you need a turnkey service or a mature Python workflow: the beta Python version has known limitations, and Robyn's strongest use case assumes substantial data and modeling capability.