I found some great background on forecasting concepts and also some good suggestions in the white paper “Demand Planning:How to reduce the risk and impact of inaccurate demand forecasts“. Regarding the suggestion of using collaborative forecasting techniques to improve the accuracy of forecasts, this is something that I’ve seen with the customers I’ve worked with (mostly in the high tech/electronics sector). Most companies of any size do some type of collaboration to develop a forecast. At a bare minimum, sales forecast data is collected on a monthly basis and marketing and product management are consulted regarding new product introductions, product end-of-lifes, promotions, discounting, competitor activity and industry trends. As the author points out, one of the real challenges related to this activity is trying to minimize the time collecting and rationalizing the data as this takes time away from the type of analysis that adds real value. This can usually be solved by choosing a suitable, scalable tool that all users will accept (which, as the author points out, is not Excel).
It’s also important to measure forecast accuracy properly. On a recent visit, I noticed that once of our customers is using a model that looks at a weighted composite of bias (actuals consistently higher or lower than forecast), error (the size of the delta between actuals and forecast) and stability (how much is the forecast changing over time) which should be applied separately to each of the collaborative inputs. Generally, collaborating with customers on forecasting is something that is considered to be difficult to pull off. As the paper points out, efforts in this area should be focused on your small set of ‘A’ customers. Also, if attempting this, you need to be sure you have a forecasting model you believe in and that has shown to work because you’ll often be pitching this process to a customer that is still in the dark ages of forecasting.
Regarding the last point about forecasting more, the theory may makes sense but I think in practice, it would be very difficult to pull off. Forecasting is inherently a process of gathering multiple points of information and finding patterns that, when compelling enough, cause a change to occur. I think rather the key is to communicate changes to the forecast to all affected parties as soon as they occur and only when they occur rather than necessarily forecasting more often.
TrackBack
• Digg This
• Add to del.icio.us
Tags: Demand planning, Forecasting
Posted in Demand management
You can leave a response, or trackback from your own site.
I have been practicing Supply Management for way too many years and for quite a few companies with differing demand needs albeit all in the Agriculture, construction and over the road heavy equipment markets. I have for the last 20 years employed every technique discussed in this white paper with varying degrees of success individually, however when all of these techniques are properly used in conjunction with one another, tied together with supply partnership agreements of some sort with collaborative expectations clearly outlined you can develop a very high performing responsive, flexible relatively lean and responsive demand system and supply chain.
Again depending on the product being produced and market supplied, it all starts with a Sales or marketing forecast which is obviously a collaboration with dealers, customers etc. The trouble with most consultants, analysts, programmers and academics alike is that they try to reduce the theory behind these techniques into mathematical algorithms which tends to suggest that implementing these techniques is more complex than need be hence supporting the theory that some sophisticated ERP/Forecasting system is required. All that’s required is a basic MRP system; the rest can be easily managed with manual procedures and processes. Granted it takes time to set things up initially but when that is done the reforecast can be done and implemented very fast. And if Sales can not get a new forecast frequently enough, a good Materials guy can adjust the whole system by forecast error from the prior month projecting into the future without a whole lot of hassle.
The basics are as follows:
-Sales/Dealer forecast by model, customer collaboration
-Sales and Materials corroborates forecast with historical/seasonal trends and projections, internal collaboration
-Develop and load master plan (previously set up as a two level plan for Mfg planning bills and pass through parts planning bills as well as any long lead buffer plan required)
-Run MRP
-Use MRP plan to firm order manufacturing and low volume option oriented purchased items and where suppliers will not support VMI/Kan Ban type replenishment systems
-Frequently/monthly collaborate with VMI/Kan ban suppliers as to where the forecast/plan is headed
-Daily/weekly VMI/Kan Ban supplier delivers parts to actual usage and or minn/max targets previously set up
-Propagated forecasts and or safety stock systems should have been previously developed for independent demand items i.e. after market service parts etc.
-Re-plan weekly or monthly based on prior months forecast error or new Sales forecast
When the procedures are set up this whole process can go rather quickly. The great equalizer in the whole process is the VMI/Kan Ban replenishment process which if set up correctly emulates actual demand throughout the system. As has been stated in the white paper forecast are inherently inaccurate so any plan and hard orders generated from MRP forecast will follow that forecast error while VMI/Kan Ban will tend follow actual demand thus helping to eliminate overages or shortages.
With the above in mind, it becomes important to consolidate suppliers, develop supply agreements, and drive flexible, short lead time, quick set up and Kan Ban type flow techniques into the manufacturing system and the whole supply base.
For a mid size company with existing Product Structure, MRP/Inventory management systems this can all be implemented in less than two years after accounting for management and employee resistence.
The impact of accurate forecasting cannot be realized until it is tied with Inventory Levels and Service fulfillment requirements of the supply chain. Many a times, accurate forecasts lie on the planner’s desktop and inventories and orders are still decided in a disparate manner. The magic triangle of “Forecasting Accuracy – Optimized Inventory Levels – Highest Service Levels” can be achieved only through tight integration of supply planning with accurate forecasting. During recessionary times, faster demand sensing plays pivotal role in preparing organizations to respond to changing consumer purchase behavior. Although, frequent forecasting (weekly or daily) is now a reality thanks to powerful processors, supply chain managers are often found confused as how they can attain the benefits of it. Linking service levels with inventory levels based on consumption forecast using simple maths (shown here) can solve this puzzle!!