Loss Forecasting — predicting future losses through an analysis of past losses. Past loss data must usually span a sufficient number of years (5 or more) to achieve some degree of credibility. The time span is important because the most recent years' experience most closely approximates current exposure, yet the earlier years' losses have had more time to develop. The law of large numbers, exposure data, any anticipated changes in company operations or structure, inflation, workers compensation benefit changes, and any other relevant factors must be considered when forecasting losses.