Theories and models in management build on the prediction of future states and outcomes. Yet, management scholars often focus on model fit metrics, while giving less attention to prediction and forecasting methods. The underlying assumption is that predictions and forecasting are byproducts of good model fit. In this paper, we argue and find that when researchers focus solely on fitting data, they risk creating models and theories that fail to generalize to other similar datasets, limiting their forecasting, practical value, and applicability. Model fit metrics are subject to overfitting and underfitting. Overfitting compromises prediction by overlearning from the data, while underfitting sacrifices prediction by failing to capture key patterns. We apply newly developed forecasting metrics to show how forecasting metrics can complement model fit metrics, enhance model evaluation, and address model fit limitations. We emphasize the importance of using multiple metrics to comprehensively assess and improve both predictive performance and theoretical development.
