- The client was able to achieve savings of approximately $500K year-on-year by replacing their off-the-shelf model with a customized solution.
- The new model offered increased accuracy by 5–8% in comparison with the previous model.
- Customized solutions for 40+ categories were rolled out in a span of two months, speedily optimizing forecasts across geographies and product lines.
A multinational personal care brand based in North America wanted to generate long-term forecasts for 30+ categories of products across their US and Canada geographies. They wanted to replace their previous off-the-shelf forecasting tool, a black box model offering limited control over customization, with a model that would be transparent, customizable, intuitive, and which would better integrate business judgments and technical outputs. The business also wanted to examine the impact of COVID-19 on their sales and incorporate the data into their long-term category forecasts.
In addition, they wanted to build an advanced analytics tool that could address the following key questions:
Long-term Forecasts: What are the base forecasts for the next five years?
Driver Analysis: What factors drive changes in demand each year and how can assumptions regarding these drivers be made more accurate?
Simulation Dashboard: How can forecast dashboards incorporate individual judgement and simulations be created based on each input by a category manager/business executive?
TheMathCompany interviewed 15+ stakeholders and undertook extensive market research to identify all the factors influencing categories and their estimated impact. A customized product prototype was built to meet the client’s specific needs, and a top-down approach was employed to ensure the elimination of bias on data available. The impact of COVID-19 was accounted for in the model, and the magnitude and trend of the forecasts of all drivers, volume, and price were fine-tuned. The weightage of each driver on volume was accurately aligned with business expectations.
The following functionalities were incorporated into the forecasting model after extensive interviews with category managers and business executives to ensure comprehensive customization:
1) Long-term Forecasts: The base demand forecasts that were to be included in the model were carefully considered and finalized after extensive research, with 6–7 different criteria selected. The dashboard allowed for the generation of forecasts for the next five years.
2) Driver Analysis: The factors impacting and potentially driving changes in demand for each year were accounted for in the model, with drivers and their weightage optimized and aligned with specific business needs.
3) Simulation Dashboard: After considering several approaches for the simulations and building multiple customized prototypes, the final forecasting tool was custom made, with managers having the ability to input their own data and judgements into the intuitive, easy-to-use dashboard to generate long-term forecasts. These forecasts also accounted for the disruptions of the pandemic, making them highly relevant to the current business context.
The category planning and forecasting tool allowed the business to generate accurate long-term forecasts and optimize savings. This tool is scalable across geographies, with the nuances of different demographics, markets, and industries accounted for, and can be deployed across multiple product hierarchies for the business as well.