Aryma Labs using Marketing Mix Modelling and Differences-In-Differences

This paper from Aryma Labs introduces an innovative approach to proving the efficacy of Marketing Mix Modeling (MMM) using the Difference-in-Difference (DiD) technique. By comparing test and control markets, the authors demonstrate how DiD isolates the true causal effect of marketing investments, addressing concerns around natural sales trends and external factors. The study highlights both the statistical challenges (such as parallel trends and autocorrelation) and the practical benefits of applying DiD in MMM, making a strong case for its broader adoption as a reliable validation framework.

Aryma Labs—a Bengaluru-based data science startup founded in 2019—specializes in marketing mix modeling, machine learning, NLP, marketing analytics, and time-series forecasting. As an unfunded enterprise, it delivers pragmatic, ROI-focused analytical solutions to clients across industries like FMCG, e-commerce, travel, and supply chain (arymalabs.com).

Led by co-founders Venkat Raman (Statistician and Data Scientist) and Ridhima Kumar (Chief Marketing Mix Modeling Officer), Aryma Labs is dedicated to building AI-powered marketing ROI solutions that emphasize transparency and real-world impact. Their vision, “More Results, Less Hype,” reflects a focus on measurable outcomes over theoretical promises.

In this study, the team introduces a novel application of the Difference-in-Difference (DiD) technique to validate the efficacy of Marketing Mix Modeling (MMM) in practice. Using two comparable markets, they applied MMM recommendations by increasing investment in the key “hero” channels (such as ch1–ch5) by 30% in the test market (Market A), while keeping the control market (Market B) unchanged.

The results were clear:

MMM prediction suggested 550 incremental sales from the increased spend.
However, using DiD, the researchers created a counterfactual scenario showing that Market A would have achieved 2,200 sales without the intervention.
Actual sales reached 2,700, meaning the true causal impact of MMM-led interventions was 500 units.

‍

‍

By leveraging DiD, Aryma Labs proved that the observed uplift in sales was not just a result of natural growth or external factors, but directly attributable to MMM-driven marketing strategies.

The paper further addresses common statistical concerns, including the parallel trends assumption and autocorrelation, and shows how careful econometric design ensures robust results.

‍

You can check this example in our Diff-in-Diff Tool: https://sterling-diff-in-diff-tools.streamlit.app/

‍

You can contact Aryma Labs:

Website: https://arymalabs.com/

LinkedIn Page: https://www.linkedin.com/company/aryma-labs/

Venkat Raman's LinkedIn Page: https://www.linkedin.com/in/venkat-raman-analytics

‍

You can contact Sterling:

Website: https://sterlingdata.webflow.io/

LinkedIn Page: https://www.linkedin.com/company/sterlingvalue/

David Dias' LinkedIn Page: https://www.linkedin.com/in/daviddiasrodriguez/

‍