Data Scientist - Freelance
Designing and industrializing causal measurement frameworks for in-store advertising campaigns: implementing advanced matching algorithms (1:1, M:N), defining statistical best practices (significance, bias control), deploying production-grade services for campaign generation and measurement, and delivering interactive dashboards to monitor performance and ROI.
About this role
Médiaperformances is an innovative Retail Media company leveraging data and advanced technologies to optimize campaign performance and shopper engagement.
As part of the Data team, my role is to develop a measurement solution for retail media campaigns, aiming to automate and standardize performance evaluation using data science and statistical methods.
The core challenge is to build comparable exposed and control groups from store data, so that campaign performance can be measured with statistical rigor. I designed and implemented a matching algorithm based on the CP-SAT optimization solver (OR-Tools) to tackle this problem at scale.
Beyond the algorithmic work, I also focused on making the solution production-ready: automated pipelines, code quality, reproducibility, and maintainability were central concerns throughout the mission.
Key contributions
- Designed and implemented a matching algorithm to build comparable exposed and control store groups, leveraging CP-SAT optimization (OR-Tools).
- Developed automated end-to-end pipelines to measure campaign performance at scale after each campaign.
- Applied A/B testing methodologies to ensure statistically robust and reliable performance insights.
- Defined and standardized KPIs for campaign evaluation across the data team.
- Built production-grade data solutions with a focus on code quality, reproducibility, and long-term maintainability.