I have a prediction. Enter a room full of operations professionals and start talking about forecasting. The number of yawns will be correlated with the number of minutes spent discussing statistics.
A mere mention of “Mean Absolute Percentage Error” and you’ll feel the urge to open your mouth, inhaling slowly and deeply.
STOP right there. This is a STATISTICS FREE zone.
Like it or not, we humans are not developed to think clearly about probability. Our brains are hard-wired to think about cause and effect. The probabilistic world is just too fuzzy and random. We prefer solid rules that we can use to prepare for the future.
I am not denying the value of statistics. Best practice in demand planning and forecasting relies on a statistical toolbox; done right, it can be very powerful. Yet first: sales forecasting and demand planning needs a bedrock foundation. Getting the fundamentals right depends on some sturdy principles.
Here are five things you need to know about forecasting.
All Forecasts Are Wrong, Some Are Useful
A sales forecast is a model about future demand. It is an attempt to predict the future — the result will either be lucky or lousy. Even when it is lucky, it still does not represent what will actually happen (For example, customers rarely buy product in neat monthly buckets).
Forecasting is often compared to driving a car whilst looking in the rear-view mirror. The past gives a few clues about the future, but not enough to stop you from driving off a cliff.
Predicting the future is expensive and often unreliable. After investing billions in computer models, financial companies are repeatedly wrong-footed by the market. And most manufacturers don’t have anything that comes close to this computational firepower.
The solution is not complex analytical software. The answer is this: Master the present before trying to predict the future. There are signals everywhere that point to how demand is changing. Adaptive manufacturers are watching and listening closely to the way customers consume their product. Respond and adapt to these changes, and you will depend less on prediction.
A fast response to customer demand is built on changes that you can control. If it takes manufacturing days and weeks to supply a product with minutes of value-added work, there is plenty of room for improvement.
Each day cut from lead-time is a day that is cut from the effective forecasting distance. There are myriad opportunities to reduce the need for you, your suppliers and your customers to forecast demand. Reduce the total exposure to forecast error and the forecasting you need to do will get better results.
If You Can’t Measure It, You Can’t Manage It
For every ten manufacturers who complain about forecast accuracy, there’s only one, perhaps two who measure it.
To measure forecast accuracy you need definitions and data. Maybe you’re unsure how to define accuracy of the demand forecast. Maybe, like many other companies, you lack the data to measure and monitor the accuracy of your forecasting.
We often asked by clients to help them manage and control the demand planning and forecasting process. Despite the fact that MRP depends heavily on forecasting, most MRP systems do a poor job of collecting, presenting and reporting the forecasted versus actual performance.
The main reason is that MRP systems are designed around transactional data, and forecasts are rarely transactions.
Customers, sales people and distribution channels share sales forecasts by emailing (or even faxing) spreadsheets (or even –urghh – PDF files). Everyone has their own format and the data are neither organized nor connected.
Reliable reporting on forecasting performance is highly valuable. There are many patterns of error, bias and gaming that are uncovered with the right intelligence. This doesn’t need complex predictive analytics – we can generate reports in spreadsheets and build solutions without the need for additional software.
Before you get a data model to measure and control the forecasting process, you need to consider these three remaining things you should know about forecasting:
Different Products Forecast at Different Ranges
Each period of actual demand has many versions of the sales forecast. If we are forecasting October demand, then this forecast may have changed every month from January to September.
Each product requires resources and materials that have a lead-time. The effective forecasting range for a SKU can be anywhere between the longest and shortest lead-times of its dependents. Which is the most relevant? Which one do we use to manage forecast accuracy?
This requires a strategic review of the items in the bill of materials, and in some cases the resources required in manufacturing. Some manufacturing processes have long lead-times on production resources. Other products are constrained by resources for marketing and distribution.
The positioning of items you stock and those you make/buy to order will be central to this review. What commonality is there for materials shared by one or more product? Are your forecasting families grouped by their requirements or are you using some legacy product family that ignores them?
Each SKU has a forecast range that is most critical. The range should be determined by a strategic review of materials, manufacturing and purchase lead-times.
The classic book by Joe Orlicky, the pioneer of MRP, is Material Requirements Planning. It has been re-written by Carol Ptak and Chad Smith. The new concept is called Demand Driven MRP and it advocates a careful positioning of stock that is optimized for effective lead-times that arise from the manufacturing process and the bill of materials. This is all very different to the old MRP concept of inventory everywhere.
A Single-Number Mentality Will Kill Your Planning
In best practice for medium to large manufacturers, sales forecasting is delivered in the sales and operations planning process. Dick Ling is widely considered to be the father of Sales and Operations Planning.
Along with Andy Coldrick, in his 2009 paper “Breakthrough Sales & Operations Planning” he describes his experiences with over 25 years of implementing S&OP. Many companies have had successful results. However, for many companies the forecasting, sales and operations planning process has failed to properly integrate with the financial planning and strategic direction of the business.
S&OP was ignored or over-ridden by financial and business planning.
Because the operational number for the supply chain was lower in priority than the financial number, and was very often different, the S&OP meeting became the forum where supply people grumbled about forecast accuracy against their ‘single set of numbers’- the impossible dream. It was becoming apparent that getting a single number from a pre-SOP meeting where people had their own functional agenda was virtually impossible. Very often the ones who shouted the loudest got their way, but if finance did not agree the number was questionable.
Breakthrough Sales & Operations Planning, 2009, Richard Ling and Andy Coldrick
Forecasting becomes credible when it takes on a range of values and scenarios. If you expect sales and operations people to produce a sales forecast comprised of a single value for each period, you doom them to failure.
A forecast expressed as a range of values can incorporate different views of the same demand. Business and financial planning accepts that the future holds uncertainty. The business needs operations to consider a supply and financial scenario for different demand profiles.
Forecasts expressed as a range of values between upper and lower limits do not necessarily require statistical distribution functions. Just by looking at different scenarios, demand planners help the supply side of the organization quantify the implications of supplying at those limits.
Measuring forecast accuracy now becomes a process of measuring how demand falls into the target range.
If You Want People to Use a Forecasting Process, It Has To Be Useful
Forecasting rarely performs well as an administrative process. If you don’t give sales and marketing people a good reason to provide a forecast, they have little reason to ensure that the forecast is a good one.
A good forecasting process delivers value to users. It helps them sell more product, make more customers happy. It gives them visibility on demand history that helps them prioritize opportunities and position channels to face growing demand.
A sales forecasting tool that gives real-time feedback on performance is engaging to sales and marketing people. It equips every rep and account manager with aggregated business intelligence and gives them a good reason to share their information.
How can you spot a forecasting process that doesn’t properly engage sales people? Check to see when the forecasting work is performed. If sales people scramble at the end of the month to supply the forecasting number, this is a sign of bureaucratic compliance.
Real market and customer intelligence pops up all the time. If a nugget of information only influences the forecast at the end of the forecasting period, this is wasted time. Responsive manufacturers need to sense and adapt to changes in the market. Sales and marketing people need the tools to react immediately to new information and have it reflected in the forecast for operations to act.
There is a lot of opportunity to improve forecasting. None of these actions require complex statistical models or investment in planning and analytics software. When you tie together the right business processes with a simple data-driven model, great results will come.
wap.igic.com.cn consultants can help you improve your forecasting and demand planning process with a powerful, yet cost-effective spreadsheet model. Click here to talk to us and learn how to put these ideas into action.