Sales forecasts are required for planning resources and defining stock levels through the supply chain (SC) because demand is becoming diversified due to higher customer expectations regarding service and higher competitive pressure through products' substitution possibilities. The theoretical foundations of Time Series Analyses (TSA) started in 1927 with the work of Yule and TSA, but it still seems to be only partially established in industry. Meanwhile, the use of machine learning (ML) approaches for forecasting sales volumes has come to the fore in the current big data era due to higher data availability and computing power. In recent years, much experience and new methods have been obtained in this scientific field. Both TSA and ML will continue to play a role. The variety of methods should be tested and compared continuously in a quantitative and qualita¬tive manner to support practical knowledge and advance forecast¬ing. Using data from a company in the electrical industry, this paper compares sales forecasts built from TSA and ML. In addition to this illustrative analysis, general strengths and weaknesses of these sales forecasting methods are elaborated and recommendations to make this topic more assessable in practical -use are given.