Forecasting tool developed to setting daily funding to stores in a specific market.
This project resulted from a need of the accounting team in a company I previously worked at to improve their forecasting methods. In order to minimize the intervention of other departments in this daily procedure, we opted for using a self-updating tool to perform Time Series Analysis, such as the Auto Arima tool implemented in the forecast R package.
The program feeds from the previous two years of data, which during early analyses showed enough information to present seasonality and to avoid overfitting. For the sake of promoting the usage ease of the tool, it consists of only an user-fed Excel file, the R script, and a batch file which presents the user interface.
This tool went through a testing period of two months, after which the information provided by the tool presented on average an increase in accuracy of 21% compared to the previous methods of forecasting used by the team.