Project Overview
Leveraging trained Artificial Intelligence (AI) modelling to streamline HR annual operations planning based on net sales forecasts
The need
To optimize operational costs incurred at a store level, one of our Fortune 500 clients required a net sales per annum forecasting solution with five-year historical data available at the store-department level. Accuracy in forecasts defined as Mean Absolute Percentage Error (MAPE) < 5% & Proximity < 2% was crucial to the success of gauging human resource and operational needs for each store such as the number of check-out counters or in-store staff required in the stores at a given point in time. Findings through data would be analyzed and incorporated in the Annual Operating Plan for the company.
The Solution
A team of certified data scientists and qualified machine learning engineers came together to employ time series-based machine learning models to address the need at hand. Sales data from the past 5 years was used to train, evaluate, and improve the models to achieve an accuracy of above 95% in the forecasts.
Anomalies in data such as missing information and holiday-driven outliers were taken into account through meticulous feature engineering and exploratory data analysis. To ensure sustainable growth prospects, modeling was also performed to allow for scalability in the future.
The Outcome
Accurate net sales forecasting
Net sales forecasting with accuracy above 95%.
ML pipeline
A reusable ML pipeline, scalable and distributed using Azure Databricks and PySpark.
ML modeling
Successful ML modeling, including techniques like Facebook Prophet, Random Forests, XGBoost, and SARIMAX.
Expert data science and ML team
A winning team of data scientists and machine learning engineers equipped with expert scientific knowledge of data processing and ML.
Enhanced data quality and analysis
Data cleaning, preparing, and manipulation through robust feature engineering and exploratory analysis.
