A retail client wanted to measure the effectiveness of different promotions in increasing their market share.
We looked at measuring the relationship of sales of different products as it is crucial to analyze what effect of promotion on some products will have on other similar products. Sales lifecycle of products were factored in to accurately forecast sales in the absence of promotion. Together this allowed us to understand the pull forward and cannibalization effects to arrive at a more robust measure of the true impact of the promotions.
- Forecast Market sales for individual products
- Forecast Sales for product groups/categories
- Incorporate promotion attributes into each of the forecasts
- Measure halo, pull forward, cannibalization to compute item’s market share
- Measure the statistical significance of the effect of promotion on market share
TECHNIQUES, TECHNOLOGIES, TOOLS
- ARIMA, Auto-Regressive and Recurrent Neural Networks, Random Forest, Croston Model, Poisson Regression, Ensemble modeling
Our clients were able to make informed decisions about which of the promotions were not working for them thus, saving costs and allowing them to focus on the ones which were more effective.