The article discusses how to make accurate forecasts using the double seasonal Holt-Winters method. Two methods were tested: percentile error bootstrap and double seasonal block bootstrap. The percentile error bootstrap is better for data without outliers, while the double seasonal block bootstrap is more accurate for longer forecasts. The accuracy of the percentile error method decreases with longer lead times, while the double seasonal block method's accuracy improves. The prediction intervals of the percentile error method remain stable over time, while those of the double seasonal block method widen.