To determine delivery dates for customers, plan production capacities, and schedule and coordinate orders, it is of great importance for production planners to work with accurately planned Throughput Times for production orders. However, conventional estimation methods, which are based on fundamental statistical principles and expert knowledge, often fail to adequately account for the numerous factors that can lead to discrepancies between planned and actual outcomes. These discrepancies can have substantial ramifications. As production systems become more complex, the limitations of these conventional approaches have become increasingly evident. The employment of machine learning models holds promise in enhancing the precision of Throughput Time prediction. However, selecting the appropriate data is paramount for training a machine learning model. To achieve this objective, it is essential to identify the key factors that influence the prediction of planned Throughput Times. Despite its relevance, no systematic investigation has yet been conducted on these factors, especially in job shop manufacturing, where orders are processed according to the Make-to-Order principle. To address this research gap, data sets from five Small and Medium-sized Enterprises with job shop production are evaluated and key factors for determining planned Throughput Times are identified. Utilizing the Cross-Industry Standard Process for Data Mining, the characteristics of the individual steps involved in constructing a prediction model are elaborated.