Underforecast

Underforecast is a forecasting outcome in which the predicted value falls short of the actual realized value. The term is used across fields such as meteorology, economics, energy planning, and inventory management. An underforecast can be contrasted with overforecast, where predicted values exceed actual outcomes. In practice, underforecast arises from model bias toward lower estimates, limited or nonrepresentative data, failure to account for tail risks or extreme events, and conservative assumptions by forecasters or decision-makers. Human judgment, organizational incentives, and data latency can also contribute.

Impacts of underforecast include inadequate preparedness, stockouts or supply shortages, missed revenue opportunities, higher late adjustment costs, and suboptimal resource allocation. For utilities and energy systems, underforecasting demand can lead to insufficient generation capacity or constrained transmission; for weather services, it may affect public safety advisories if critical thresholds are underestimated.

Measurement and evaluation: forecast accuracy is assessed by errors, with bias measuring systematic under- or over-prediction. Common metrics include mean absolute error, root mean squared error, and mean forecast error (which indicates bias). Calibration and backtesting on historical data help identify persistent underforecasting tendencies.

Mitigation strategies: use ensemble or probabilistic forecasts to represent uncertainty, incorporate scenario planning for extreme events, continuously update models with new data, apply safety margins or buffers, and continually validate against out-of-sample data. Transparent communication of uncertainty is also emphasized to support informed decision-making.

See also: forecast bias, overforecast, forecast error, probabilistic forecasting.