Analysis of Sales Shift in Retail with Causal Impact: A Case Study at Carrefour

Estimating the sales shift from an unavailable product to other items is a complex challenge for retailers like Carrefour. This case study applies Google's Causal Impact model, which uses a Bayesian structural time-series approach to create a synthetic counterfactual of what sales would have been if the product remained available. The model's design involves a specific heuristic for selecting covariates, combining time series with high co-integration for long-term trends and high correlation for short-term variations. To validate this approach, the model's performance on partial unavailability cases is compared against both a difference-in-differences analysis and a Causal Impact model using a true control group, demonstrating its reliability for broader application.