What is Robyn?
Robyn is an open-source Marketing Mix Modeling (MMM) package developed by Meta Marketing Science. Robyn's mission according to Meta:
"Our mission is to democratise modeling knowledge, inspire the industry through innovation, reduce human bias in the modeling process & build a strong open source marketing science community."
This mission is part of Meta Marketing Science's broader goal to help businesses grow by leveraging data and science-driven marketing practices
With Robyn, skilled analysts can create Marketing Mix Models with ready-made modeling tools:
Building models
Reviewing model outputs, such as Marketing ROI, Diminishing return curves
Optimizing budget allocation
Robyn makes it faster for data scientists to build Marketing Mix Models, so that they don't need to spend time themselves implementing features such as adstock or diminishing returns.
Robyn is typically compared with Google Lightweight MMM, which is another open-source library for Marketing Mix Modeling. For a detailed comparison and review of Lightweight MMM vs. Robyn, you can check this blog post.
Who is Robyn for?
Robyn is built for
Data scientists and skilled analysts, who wish to do the whole Marketing Mix Modeling workflow themselves, from data cleaning to model outputs.
Hobbyist modelers and statisticians who wish to learn more about how Marketing Mix Modeling works
Robyn vs. Sellforte: A Comparison guide
In summary, Sellforte is an end-to-end MMM platform, while Robyn can be considered a modeling package or library.
1. Data integrations
Sellforte uses automated data connectors when connecting data. This minimizes the time spent on data update in the future, and enables using Sellforte's standard data processing pipelines that are cornerstone in delivering high quality MMM results. Below is the data connector view in the Sellforte product, that is available from the public Sellforte demo. Using this view, users can connect data to Sellforte. Sellforte also has other tools available for connecting data.
Robyn: Users collect the data from different ad platforms themselves.
2. Data processing
Context: Raw data, coming for example from ad platforms or ecom platforms, is rarely usable for Marketing Mix Model without any processing. Typical data processing activities include for example data validation, campaign mapping, and building media hierarchies.
Sellforte offers standard data processing as part of the solution. Users can also export processed data to other platforms, tools or dashboards they might have.
Robyn does not offer data processing tools, beyond basic data transformations available in R and python.
3. Modeling
Sellforte conducts modeling on behalf of the user, leveraging the modeling expertise it has built over the years. Sellforte takes responsibility of result quality. Quality in Sellforte's modeling is built on 3 pillars:
Intelligent data processing: As an output from the data processing, Sellforte's modeling pipelines start from a clean and MMM-tailored dataset. (Note: this is an overlooked topic in MMM)
Proprietary configuration of Bayesian modeling approach: Sellforte's modeling approach is based on Bayesian inference, the golden standard for MMM. Sellforte has a proprietary configuration of Bayesian modeling, and standardized model validation flow for maximum quality.
Standardized calibration methodology: Sellforte's models are calibrated with the best available information for each media (lift tests, attribution data, Sellforte benchmarks..)
Robyn enables users to conduct modeling themselves, and has an extensive documentation available (illustration below).
5. Regular model updates
Sellforte's models are updated regularly automatically as new data comes in.
Robyn's models can be updated by manually updating the data in the model.
6. Reviewing historical marketing performance
Sellforte provides a user-friendly online user-interface, that can be accessed online anytime. Historical results can be reviewed with various charts, and can be filtered in many ways. Below is a screenshot from the public Sellforte demo:
Robyn does not have a dedicated user-interface for results analysis. However, users can plot data and model outputs. While this might be sufficient for a data scientist, it is difficult for the marketing team to access MMM results with this approach. One option for the data scientist is to create reports based on the analysis. However, this often leads to a lot of extra work for the data scientist. Below is an example of a result output plot (source: Robyn's documentation):
7. Optimizing and testing scenarios
Sellforte provides a user-friendly online tool for optimisation and scenario planning. Below is a screenshot of the tool from the public Sellforte demo. With the tool, users can for example find optimal budget allocations with different budget levels. In scenario building, users can define various constraints and parameters for optimization, such as channels to optimize, and budget constraints per channel.
Robyn does not have an optimization user-interface, but the user can define optimization scenarios in code, and review optimization outputs as plotted graphs and tables. Running scenario analysis can be slow and laboursome with this approach, as the data scientist has to translate marketing team's questions into code for the Robyn library, and then communicate the outputs back to team. Below is an example of an optimization plot (source: Robyn's documentation):
8. Making conclusions based on MMM
Sellforte provides Customer Success service to its customer for interpreting the results and drawing conclusions based on them (support level depends on the contract agreed between Sellforte and customer).
Further reading