A leading consumer goods companies with brands in hair care, skin care, edible oils, health foods, male grooming, and fabric care categories was achieving low accuracy in forecasting demand across its product portfolio.

Client had developed simple rules-based demand forecasting models with less than 70% accuracy on average resulting in:
❖ Frequent stock outs of some SKUs during high demand months
❖ Excess inventory of SKUs having low demand

Client wanted to build predictive models that forecast sales of different SKUs across all its brands and in top sales areas at a retailer level to optimize inventory.


TransOrg developed machine learning based demand forecasting models that predicted daily demand at a ‘product-SKU’ pair level across the chosen ASM areas.

In phase 1 of the project, Client shared three combinations of brand and areas/zones (ASM):
❖ Brand 1 for ASM areas ABC1, ABC2 and ABC3
❖ Brand 2 for ASM areas PQR1 and PQR2
❖ Brand 3 for ASM areas UVW1 and UVW2

TransOrg analysed three years of historical sales data at an SKU and area level and developed multiple predictive models to forecast the demand.
Over 150 secondary variables for each SKU, area and month combination were developed to train the models.

The model built using random forest algorithm performed best with an overall accuracy of over 85% accuracy for the selected brands. These daily forecasts are then used by the client to optimize its inventory.


TransOrg Analytics helped the client to predict sales and demand with an accuracy of 85% which helped our client in improving inventory planning and avoid stock out situations.

Want to learn more about TransOrg’s value proposition, solution methodology and implementation approach?