iFood Case: CRM Campaign Optimizer
Business
Machine Learning
iFood data analyst case: customer analysis and optimization of a CRM marketing campaign.
Problem
Increase the return of a marketing campaign by targeting offers to the customers most likely to respond.
Approach
Exploratory analysis of customer profiles and modeling to prioritize who should receive the campaign.
Technologies
Python pandas scikit-learn CRM Logistic Regression K-Means SHAP
Results
A carefully preprocessed Logistic Regression outperformed XGBoost (ROC-AUC 0.933, PR-AUC 0.737, 92% recall). With the threshold chosen for profit (0.45), the campaign goes from −3,046 MU to +279 MU, capturing 90% of the maximum possible profit. K-Means segmentation identified a Premium group with a 31.9% response rate.