1. Time series model for sales predictions in the wholesale industry
- Author
-
Drazen Orescanin, Ana-Marija Petric, Tomislav Hlupic, and Skala, Karolj
- Subjects
010302 applied physics ,Operations research ,Process (engineering) ,02 engineering and technology ,01 natural sciences ,Data warehouse ,Product (business) ,Order (business) ,Data mart ,0103 physical sciences ,0202 electrical engineering, electronic engineering, information engineering ,Revenue ,020201 artificial intelligence & image processing ,Business ,Autoregressive integrated moving average ,Time series ,forecasting models ,sales predictions ,time series ,wholesale industry - Abstract
The prediction process in sales is a basis for a successful ongoing planning process for any organization. Wholesale companies, being B2B oriented, have to plan their organisational environment carefully to optimize the costs and maximize revenue. As the sales process is intersected with logistics, having precise sales predictions optimizes both sales and logistics processes. In order to track the sales towards a customer, we propose a data mart built on the top of the data warehouse to be used with daily loads of outgoing invoices and uninvoiced shipments data.Predictions are based on ARIMA model, one of the most popular forecasting models for the time series. The data is aggregated on a weekly level, as it was proven to be the most useful in this process. For the prediction purposes, we are focusing only on the outgoing invoices. From the business perspective, each product is tracked with data about the sales market, customer, quantity, and the date. In the article, the process of data preparation will also be included as it is the crucial step for successful prediction.
- Published
- 2020