Considering the fact that forecasting is the foundation for many business processes, it is surprising that so many companies still struggle to create a robust, value adding process against which to plan their supply chain operations.
Traditionally, the focus of forecasting has been on the technical aspects, with forecasters valued for their techniques and processes. However in the wider business, forecasting is often a heavily politicised and damaging process, with blame for inaccuracies being directed at forecasters, leading them to focus on avoiding ‘being wrong’ rather than striving to maximise their contribution to the business.
It is of course possible to not forecast at all and simply react to customer requirements. This is exactly the philosophy that underpins the Kanban approach, where supplier replenishment orders are directly linked to customer demand, with buffer stock held to deal with the volatility of demand until the replenishment order is received. This is equivalent to using the previous period’s actual sales as next month’s forecast, something known as a naïve forecast.
That said, this approach is not an effective solution for the majority of businesses because the volatility of sales would require high levels of working capital – dead cash being tied up in stock.
For forecasters to add value they must be able to anticipate this demand volatility to reduce the burden of buffer stock, while maintaining the same level of customer service. Amazingly though, in our experience, 30-50% product forecasts are typically found to be value destroying, i.e. worse than a naïve forecast.
The reason for this dire state of affairs, and companies’ ignorance of it, is that they focus on traditional metrics such as forecast accuracy and MAPE (a measure of prediction accuracy of a forecasting method). The problem with these measures is that they are only really useful at indicating whether forecast performance for an individual forecast is getting better or worse over time, not how good it is. This is because they do not adjust for the unpredictability of the underlying demand. So forecasters who boast of incredibly low levels of MAPE are usually forecasting very steady or predictable demand, which is very easy to forecast.
The key is to understand the ‘forecastability’ of individual products – the more volatile the data series, the more difficult it is to forecast. By comparing the actual forecast error to the simple naïve forecast error and the theoretical maximum level of contribution that forecasters can reach, it is possible to understand how much value the forecaster is actually adding. This allows the team to understand which forecasts can be improved and the best way to achieve this.
The challenge of implementing this simple approach is that for the measures to be meaningful they need to be calculated at the very granular level at which replenishment orders are generated over a range of periods. This can be burdensome, requiring considerably more horsepower than a spreadsheet. As a result, most companies remain ignorant of the value that their forecasting process is adding.
As a result they fail to grasp how by understanding where specially their forecasting process is adding value, they can rapidly and dramatically increase their overall forecast accuracy by addressing those markets, customers or demand planners responsible for value destroying forecasts.
By getting this fundamental of demand management right, virtually every business will see an improvement in their forecast accuracy and forecasters will make a decisive contribution to the value of the business.