Our client, a US-based private equity firm, wanted us to analyze sales data for a portfolio company in the retail sector.
The current planning process for promotional and non-promotional seasons relies heavily on old, classical business rules developed over time. There was a need to realign these rules with current market trends and develop predictive models in order to have a robust demand forecasting process that accurately accounts for promotional impact.
JMI Implementation
JMI Collected sales and promotional data across stores and blended them to have a uniform item-by-store-by-day level of granularity.
Standardized data is obtained by pre-processing data to de-trend, de-seasonalize, and remove holiday effects.
Built forecast models for baseline and promotional estimations, using a combination of time series and a frequency-based approach.
Modelled separately for regional Black Friday-type promotional events and identify the most selling and profitable categories.
Used store-level daily sales distribution indices to break down weekly to daily sales.
Identified similar products for new products by clustering for 10+ features (like sales, regional preference, complement, substitute, promoted group, etc.) and modelling based on them.
Automated the end-to-end process and enabled calibration for daily or weekly updates.
Results
Our forecasting models assisted the retailer in generating items, store, and day-level forecast accuracy improvements ranging from 5% to 60%, with a weighted average improvement of around 20%.
Estimated business value due to stock-out avoidance and improved product mix was around a 1.5% to 2% net increase in overall sales.
