Roughly Right or Exactly Wrong? Why are we still forecasting? For many companies, forecasting is a relentlessly difficult if not impossible task, it takes a significant amount of time and is often inaccurate. It will not surprise you that the Covid-19 situation has made this task even more difficult. Often pressure is placed on the forecasters to improve Service Levels whilst at the same time minimising stock by simply forecasting better. The lack of a crystal ball makes this quite the challenge. Plus, traditional metrics such as MAPE (Mean Absolute Percentage Error) rarely provide insight into the reasons of the current forecast performance. So, what to do?
R&G takes a different approach and has consistently shown to lead to more positive results by leveraging all the information stored in your own data and placing more responsibility with the entire supply chain.
Below are some thoughts you might want to consider:
The main problem with MAPE is that no learning takes place. You can only look back to see if there has been an improvement in forecasting, the information on why this change happened has already been lost in the calculation! Additionally, by working with the absolute value, the forecast performance span is greatly reduced. This is the percent point difference of the weeks with the largest percentage of UNDER-forecasting and the weeks with the largest percentage of OVER-forecasting. As a result, a good analysis is no longer possible.
Furthermore, the forecast error is unfairly mitigated by working with the average of a group of items. The over-forecasted items thus compensate for the under-forecasted items. However, the customer experiences it very differently. For them, a stock-out on one item is not compensated with an excess of another item. Therefore, it is much better to look at the various items independently at a transactional level. If you do this with a multi-functional team, it will soon become apparent that it is not solely down to the planner why forecast errors can arise.
Be smart on your historic data. For many companies, forecast error is a fact of life as demand is often volatile and difficult to predict. Increasing the pressure on the planner to improve service levels is often not the best method. In practice, we regularly see that a replenishment strategy (for MTS items) based on statistical analysis of the historical demand works better. With this we see service levels improve whilst driving stock levels down. In some cases, you can further reduce the out-of-stock risk with special actions in case of predicted sudden surges in demand.
The use of historical data does have restrictions, for example, with new product introductions, large customers on-boarding or leaving, or super irregular demand, you cannot rely on the past. None of the forecasting methods work in these situations.
Evidence shows that the R&G Stable Supply Chain method continues to produce better results. If the above has triggered your interest and you would like to know more, please feel free to contact us.
Bart Beusmans is Business Process Consultant at R&G Global Consultants, located in the Netherlands.