Amcor Flexibles North America (AFNA), a manufacturer of packaging solutions for a variety of industries, often faces challenges in ensuring the on-time fulfillment of their orders, as most large companies do. Order lateness is difficult to identify proactively, and brings with it negative repercussions such as reduced customer satisfaction, sales declines, and cost increases. This project aimed to develop a machine learning model capable of making predictions about whether AFNA’s individual sales orders would be delivered on time, to enable opportunities to mitigate it, or give the customer earlier awareness of the delay. Additionally, the project sought to gain insights into order characteristics that most significantly affected order lateness, to allow for strategic improvements. Various data was gathered and used to train a suite of machine-learning algorithms, to identify a single model that reported the greatest F1 score. This produced a random forest model that exhibited high predictive strength, correctly classifying over 90% of sales orders. In addition to providing binary classification, the favorable distribution of predicted probabilities could be classified into smaller groups that could receive different degrees of mitigation effort, minimizing costs associated stemming from misclassified orders. Gini Importance revealed that material complexity and production location played a large role in determining order lateness than previously anticipated.