Working as a supply chain consultant for Enchange primarily in FMCG and Pharmaceutical companies, S&OP is probably the major area of project work. In these companies forecasting varies from a "good guess" to the use of the most sophisticated IT tools using mathematically designed algorithms and it has to be said that the accuracy gained from forecasting methods from both ends leaves much to be desired. So why can't we forecast accurately?
One favourite excuse is that the low level of maturity in the market provides a level of instability that does not allow the supplier to be able to forecast to a level of accuracy required. If customers don't know what they are going to sell in the month how can we be expected to forecast accurately? Other excuses stem from the dichotomy between supply chain "push" and "pull".
We were involved recently with a maker of specialist bathroom fittings that has a captive market and sells all it can make and therefore says why I should bother to forecast!
A pharmaceutical company that has an "ABC" product classification has stopped forecasting for all "A" products and used a replenishment method of managing factory finished goods inventory against set levels indicated by Green, Amber & Red values. When stocks of a product fall from the Green level to the Amber level a production order is raised and before falling into the Red level the order is being produced. However, for most companies we have to forecast and must forecast accurately to provide expected levels of customer service and manageable supply chain costs.
So how do we do this and provide an accurate input to our S&OP process? Well here goes - firstly we need to agree not to forecast all products - it's just not worth it so let's stick to the 20% of products that provide the 80% of our business (turnover & profit). Secondly let's consider all the internal and external inputs we can to the forecasting process. A recent brainstorming session with a client identified a possible 20 inputs which was up from the previous 8 inputs. For all inputs we need to verify the accuracy of information and the reliability of the person or organisation providing them. Thirdly not all inputs have the same importance we need to weight the inputs as to their significance to the volumes being forecasted. We now have a recipe to make the forecast and what we need to do now is to throw all the ingredients in to the pot (database) and see what comes out. If you get "gobbledegook" you will need to check the accuracy of your inputs and weighting. A good rule of thumb is to look at the "this month last year volumes" and the sales trends this year compared with last year.
Is forecasting a science or an art? Well it's a bit of both with what I think is a higher weighting towards science. What level of forecasting accuracy is achievable for my company? Well it's the value that provides the optimum levels of customer service you require at the supply chain costs you can manage.
