Forecasting Customer Demand
Preparation for the end of year sales season of Black Friday, Christmas and New Year means that a fashion business needs to estimate demand for SKUs of clothing by size and style. In previous years the approach has been to take an educated guess on what fashion styles will do well and order a range of sizes to fit most body shapes. This usually results in a considerable quantity of unsold items that later need to be disposed of at a steep discount or loss.
Forecast demand based upon past sales history using machine learning forecasting models that can forecast demand including for new products.
Photo by Ussama Azam on Unsplash
Machine Learning to the Rescue
Use past sales data to extract qualified time series data for products into a data lake. As long as you have a time series data set of past sales we can import the data into Amazon Forecast, prepare the forecasting model with any required parameters and generate Forecast percentiles. Using the Forecast data the buyers can choose how much above a mean value to use to ensure sufficient supply without excessive unsold stock. Machine learning models are not perfect but give objective guidance to assist purchasing decisions. If required the whole process can be automated using new sales data on a defined time period by defining automated ETL processes with AWS Glue driving the Forecast ingestion and processing to feed into Amazon QuickSight for business intelligence visualisation.
Photo by Fezbot2000 on Unsplash
Use our Contact page if you would like to learn more.