Looking at the chart below, demand management can be broken down into 3 parts, namely, the known, the unknown and the really unknown. Let’s look at each of these separately.
The Known is the easiest part as this is basically your customer order book. You know exactly what is required and when. Unfortunately, in most businesses these orders only extend to the next few days or weeks, if you are lucky. We just need to make sure that these orders are entered onto your system as soon as possible such that they can consume the forecast, if normal demand, or get added to the forecast if they are abnormal demand. Normal demand orders are those that you were expecting and were forecasted, abnormal demand orders are those that you weren’t expecting and came out of the blue and should therefore be added to the demand.
Things start getting a bit more difficult in this area. Here we have to forecast current customers and current products. This is where we use quantitative statistical forecasting to assist us in looking for trends, patterns and seasonality and projecting them in the future to give us some indication of what the future might hold. It is at this point that most salesman get a spreadsheet of numbers and they are asked to project the future and generally the best that is achieved is a simple moving average out into future. Simple moving averages are fine for stable, non-trended, non-seasonal products, and I would suggest that you probably do not have too many of those types of products. Hopefully, your product sales are growing and more often than not demand is not constant on a monthly basis and exhibits some form of seasonality.
Let’s look at some real data here and the results you would get using a simple moving average and then doing the job properly.
In the above graph, the green line to left is the sales history for a product going back about 4 years, the level red line to the right is a 3-month simple moving average forecast. The blue lines to the right indicate the confidence the system has in that forecast and basically represent +/- 3 standard deviations from the projected forecast. The red line to the left trying to follow the green sales history, known as the fitted value, is what the system would have forecast using a 3 month moving average. Moving averages always lag the actual demand which can be clearly seen in the diagram. The Mean Absolute Percentage Error (MAPE) for this example was calculated at 41%.
Now, if we ask the expert system to look for an algorithm that gives us the least errors we get the result in the following graph. In this case the system has chosen multiplicative Winter’s method and the MAPE is now nearly 74%, a significant improvement is forecast accuracy, literally in seconds, just at the press of a button, with no hassle. Now you can see the forecast has picked up the seasonality of this product and the confidence in the forecast is greatly improved. In addition, the fitted value follows the seasonality of the sales history fairly closely.
Now all we have done so far is to look for patterns, trends and seasonality and project them into the future. But, we know the future is not a direct representation of the past, so we need to collaborate these forecasts. This means talking to your customers and getting the salesman to adjust the forecasts at the detail level. These adjusted forecasts will then be used by the demand management team, to generate the consensus demand plan.
The Really Unknow
Now for the difficult bit, the really unknown demand. This is where we try and determine the demand for new products and new customers as well as consider the business environmental factors that affect demand, over which you have no control.
If you have a new product/market development function, they will be required to supply information on new products and customers along with launch dates and expected sales. If you are preparing quotes for you current and prospective customers get sales and marketing to estimate the probabilities of obtaining that business and develop a way of including this in your demand plans.
All businesses work in an economic and business environment over which they have no control. Things like interest rates, car sales, the weather, Forex rates, the oil price, commodity prices, cost of power, recession and boom periods, major sporting events, etc., etc., etc., influence your demand and you really need to understand which of these affect your business and in what way. These effects would probably then be applied to our demand plans at the product family aggregate level to come up with a more realistic bigger picture demand plan.
CFPIM, CSCP, SCOR-P, CPF, CS&OP, PLS, CDDP, CSCA, CDDL, CLTD, DDPP, DDLP, AEF, CSSC, CPIA Chief Executive Officer at Kent Outsourcing Services
Email : email@example.com
LinkedIn - Profile