technique (RMSE). According to this technique, the difference between the actual value and the forecast value is computed and then squared to standardise positive and negative errors. The square root of the average is then computed: the larger the RMSE, the less accurate the forecast. Of course, only forecasts with the same time horizon have to be used in the technique: hence, we will get three series and three RMSE, corresponding respectively to the one, two and three year forecasts.
At this stage, RMSE for this firm can be compared to RMSE for other firms or, in absolute terms, with the performance of simple random walk technique, where the forecast value of a variable is assumed to be the same as its actual value this year. If RMSE for random walk is lower than RMSE computed by the firm, perhaps the services of this stockbrocking firm should be discontinued!