Problem definition. Operations and decision making, often rely on accurate forecasts. The business forecasting process requires choosing the appropriate forecasting method, evaluating forecasts, aiding expert interventions, among other activities. In such activities, the metric we select to assess the forecast performance becomes crucial. These metrics are often used as proxies for the economic impact of supported decisions that is much harder to quantify. However, this connection has not been adequately investigated. With forecasts, we are interested in the magnitude (accuracy) and the direction (bias) of their errors. Although the first has been extensively researched, there is very limited work on bias indicators. More importantly, growing evidence relates forecast bias with the quality of decisions, for example in inventory management. Yet, we lack metrics that are informative for single or multiple items, reliable irrespective of data properties, and intuitive to the user. Lacking appropriate metrics can result in poor forecasts with adverse results on operations. Methodology/results. We propose a new metric that explicitly captures both accuracy and bias. We explore its connection with inventory management, using a real case study. We reaffirm the importance of forecast bias for inventory performance and demonstrate that the proposed ``bias coefficient'' exhibits the strongest connection from other existing alternatives. Managerial implications. The proposed metrics can help analysts evaluate forecasts directly while retaining a strong connection to operational decisions, driving improvements in the process. To support this, we provide novel visualisations based on the geometric properties of the new metric, which can aid all modellers, users, and decision-makers.