COVID lockdowns impacted us like never before in our living memory. From a supply chain standpoint, it messed up the sales pattern, left a devastated trail of bad sales history and created pent up demand spikes that further added to the mess. These distorted sales will be the basis for forecasting in the near future and given the fact that identification of good seasonal pattern, requires 2 to 3 years of historical sales data, the impact of the lockdowns is not going to leave us anytime soon as some part of the distorted data will always be a part of the forecast.

It is therefore prudent to treat the impacted data as such and limit the impact of the distortion on future forecasts. One way is to correct the history manually. This is a manual time-consuming exercise and lot of changes will be required in terms of reporting and KPIs as the base data field will change.

A better way is to have the COVID correction functionality embedded in the algorithm. Using advanced statistical algorithms, the impacted data can be corrected on the fly and the resulting forecast will be much accurate.

COVID correction technique is very different from the standard data correction techniques like outlier correction, missing data and negative value correction etc. and should be treated as such.

Based on actual sales data of many customers, it was found that compared to forecasting without COVID correction, the accuracy improved by 12% – 20% when the COVID correction parameter was enabled. This parameter therefore should become the de-facto standard in any good demand planning solution.