To date, the company’s earnings before interest and taxes have been difficult to predict because the number of orders processed does not imply how high the effective monthly profit will be. Consequently, the management can only predict the profits based on expert knowledge and personal assumptions. The challenge for management, therefore, is to improve the profitability of the business through internal optimization of processes.

The forecasting model is based on the income statements of 29 consecutive months. The model was tested in two approaches: first, by a machine learning algorithm called LightGBM, and second, by a performed regression analysis. In the first step, the results of the LightGBM model were calculated with the support of JupyterLab Notebook and Python programming language. The second approach is an Excel-based model mainly based on regression analysis on total sales, total variable costs, total fixed costs, and total costs.

The results showed that the regression analysis exceeded the predictions of the algorithm. In addition, the final formula for calculating EBIT was determined. The search for the most accurate predictions also revealed weaknesses in the revenue recognition process that make it difficult to evaluate actual revenue. In addition, it was found that improving the interfaces between departments could improve the transparency, informativeness and predictability of cost and revenue streams. Based on the results, it is recommended that the EBIT equation be further analyzed, and the optimal values of the different variables be investigated. In addition, it is suggested that the company focus on improving the data interfaces within the departments in order to bring more transparency to the revenue and cost streams and consequently have more influence on the effective costs.