Announcing the Python version of Project Robyn

Join beta testing for the Robyn Python package.

We are thrilled to announce that the native Python package for Robyn (the top request from our open-source community) is ready for beta testing. You can be among the first to try it here today - access the code on PyPI or GitHub.

A native Python version will accelerate the adoption of Robyn by the Python community and, as such, further simplify the implementation of high-quality marketing mix modeling in the industry. The R-to-Python porting was completed leveraging the latest Large Language Models (like Llama) and open source AI frameworks, demonstrating both the power of AI and our ongoing commitment to equip entrepreneurs and advertisers with cutting-edge, best-in-class measurement tools.

A look back at Robyn

Since its official release over three years ago in November 2021, Project Robyn has inspired tens of thousands of users and businesses to experiment with a modeling technique called marketing mix modeling (MMM). This has fueled a new generation of advanced SaaS MMM measurement providers as well as scalable in-house models (see Deloitte's 2024 in-house implementation framework for more).

As MMM continues to evolve within the open-source realm, companies gain deeper insights into their marketing investments, enabling them to spend smarter, create more impactful campaigns and grow their businesses.

Over the last few years, we have seen hundreds of success stories abound from all over the world, from specialized agencies like Kraz in Spain, Echo Marketing in South Korea, Yotta by Publicis in Poland and many more, along with advertisers such as Bark, Bold, Glint, LPP, Parmalat, Philips Domestic Appliances, Sneaks Up, Spoleto, Talisa, Tod’s or others.

Open innovation & community

We are committed to solving the hardest challenges in ads and marketing measurement. With features like multi-objective optimization, saturation calibration with reach & frequency and budget allocator with revenue or ROAS targets, Robyn has inspired new avenues of modeling philosophy. The first-movers and edges of the measurement industry have evolved the focus towards more business-relevant and rigorous topics like parameter identification, calibration, use of reach & frequency data and marginal effects. We’re proud to have contributed to these trends.

It is our commitment to keep innovation accessible to everyone. Operating under the MIT open-source license, Robyn is truly free and allows for monetization and commercial use. Our transparency and investment to the democratization of MMM are appreciated by the open source community and the industry at large.

We are honored by the strong community engagement from across the measurement industry. Our active forum on the Facebook group hosts over 3000 followers and the total Robyn R-library download has surpassed 70,000 organically on C-RAN alone, accounting for 40% year-on-year growth.

The community has been instrumental in shaping our understanding of application use cases and product road maps, exemplified by the community-driven release of the Robyn API wrapper - a solution for accessing the Robyn R library from other programming languages. In addition, multiple contributors outside of Meta have directly contributed to the package.

The road ahead

Thanks to all these recent innovations, marketing mix modelling is now more efficient, and dynamic than ever before. Meta support all advertisers’ ability to use MMM in three key ways:

• Data access: We facilitate granular data access to Meta campaigns for MMM through self-service & Insights API.
• Partnerships: The measurement providers and agencies play a vital role in setting accountability and ensuring the optimal return on investment for the entire media industry. We continue to partner with and support third-party MMM providers via a growing ecosystem of Meta badged MMM partners.
• Research: We’re committed to ongoing collaboration across the industry and via open source solutions to drive relevant and incremental improvements to MMM methodologies.

Looking at research specifically, we believe that several areas hold the key to addressing the most pressing challenges. These key areas include:

• Faster, deeper & integrated Models: Advertisers need timely and granular insights with frequent updates, integrating seamlessly into automated media buying systems.
• Cross-channel marginal equilibrium: Achieving optimal media budget allocation across channels requires understanding media interactions and cannibalisations across all sales channels.
• Triangulation of measurement: Effective measurement systems integrate and calibrate between experiments, attribution and modeling, addressing all model estimates including effect size, saturation and ad-stock.
• Long-term effect: A comprehensive modeling framework should incorporate full funnel logic, reflect customer lifetime values and baseline interactions.

We're particularly interested in finding ways to better harness the collective intelligence of the community to collaboratively build this future. We believe that by working together as a data science community, we can overcome these challenges and continue to innovate in the industry.

Robyn remains a truly open-source initiative, and we're committed to maintaining this approach moving forward. We're excited about continuing our journey of innovation and transparency in the industry, and we look forward to exploring these opportunities with our community. Join the beta testing for Robyn's Python and access the code today - access the code on PyPI or GitHub.