The Forecast Was Wrong Again!


By Don Rice


            Another weekend at the beach washed away in the rain. And the weatherman swore the weather would be perfect! Ski trips that are too warm; picnics that are too cold. The only thing we can agree on is that weather forecasts are more often miss than hit. There’s another forecast that’s always wrong. The sales forecast. In fact there are people in some businesses who’ll tell you that, if anything weather forecasting is more accurate than sales forecasting!       Let’s take a quick snapshot, for example, of a chart showing our actual customer orders versus our forecast...











             No surprise here. The sales forecast was wrong again. We’ve been telling sales and marketing to get the forecast right for years, but to no avail.

            When the sales forecast is wrong, it can cause a host of resource-gobbling problems, from late customer deliveries to increased inventories and higher costs. Sometimes it seems that all our business problems are caused by bad forecasts.

            But suppose we figure out a way to get around the forecasting problems? What would the benefits be?

            The answer is “substantial!”

            In fact, companies that have overcome the forecast hurdle have seen numbers like:

            <          30% increases in on-time deliveries

            <          50% reductions in inventory

            <          a staggering 80% reduction in customer lead times

            <          20% reduction in total business costs

            How can we achieve those kinds of numbers?

            First we’ve got to get back to basics--supply and demand.


Supply and Demand

            After we sort through all the peripheral issues, everyone agrees that the business objective is to make what the customer is buying. Simply stated we want supply to equal demand. This gets challenging when we have to forecast what the customer wants to buy because a forecast is really only a “best guess.” If a bad forecast is the root of most business problems, then all that’s needed is an accurate forecast.


Easier said than done?

            If the solution to all our manufacturing problems is getting an accurate sales forecast, then we have a rough road ahead of us! A forecast is, by definition, always going to be “wrong.” How much time and money have been spent on “getting a better forecast.” Lots. We visited one factory where the general manager told us they’d completely abandoned their traditional sales forecasting plans and resorted to voodoo to figure out what they were selling the next quarter. The sales people laughed, but not all that much.

            Have we gotten better sales forecasts? Somewhat, but it’s still a major problem area. But sales forecasting doesn’t have to be a source of serious problems. In fact, we can get around the issue of “bad forecasts” by going back to the supply and demand basics. The real issue is not the forecast, but rather how to reduce the variability in both demand and supply processes.

            Let’s look at six issues that affect a company’s ability to improve the supply and demand processes:

            <          single responsibility for forecasting

            <          not identifying all components of supply and demand

            <          intentionally excluding pieces of the demand process

            <          not managing demand variability

            <          supply and demand performance measurements that drive independent processes

*       supply and demand managed independently


Forecast Responsibility

             Traditionally, we have assigned responsibility for forecasting to the sales and marketing groups, even though the users of the forecast include most of the other business functions. The result is that everyone involved in the demand (and supply) planning process doesn’t have ownership in that process. When the forecast is “bad”, sales and marketing bear the brunt of the blame. The other functional areas lose faith in the forecast (does that sound familiar?) And begin creating their own forecasts.

            The first step toward correcting this problem is accepting that the forecast is actually a company-wide document, then establishing formal lines of communication between departments to guarantee that the forecast is just that. Secondly, a sales planning process, of which one component is sales forecasting, is critical to the communication linkage.


Identifying Supply and Demand Components

            In an ideal world, we manufacture what the customer wants in the exact quantity the customer needs. In fact, the most basic manufacturing equation is simply:


Text Box: Supply = Demand =Demand







If the demand goes up, you make more. If the demand goes down, you make less. The past twenty years, however, have changed both sides of the formula. The size of the manufacturing world has changed. No longer is it a regional or national economy. Competition is coming from the other side of both oceans and all of our borders. Global supply often exceeds global demand.


Compare this to the 1960's and 1970's when you could almost guarantee selling all you could produce. The challenge now is to make only what you can sell and not to make all you can. Consequently, predicting demand has become more critical, and balancing supply and demand has become a critical chore.


Let’s look at the demand side first. Everyone works hard to sell as much as possible. Expected sales, i.e. customer orders not yet received, are forecast. Customers place orders for the products that are shipped from finished goods inventory or produced to order. The demand side of the equation includes some or all of the following:

            1) Forecast

            2) Customer orders

            3) Changes in customer order backlog

            4) Changes in finished goods or semi-finished inventory

            5) Distribution or warehouse requirements

            6) Spares or service parts requirements

            7) Inter-plant requirements

            8) Samples, demos, trials, etc.


On the other side, we have all the various resources of a factory that exist to allow people to make a product:

            1) People

            2) Equipment

            3) Buildings

            4) Material

            5) Supplier capacity


Our simple Supply=Demand chart now looks like this:


To control the supply and demand components, we must first identify all the components.


Supply                 =                Demand


1) People                                        1) Forecast

2) Equipment                                      2) Customer orders

3) Buildings                                     3) Changes in customer order backlog

4) Material                                      4) Changes in finished goods or semi-finished inventory

5) Supplier Capacity                            5) Distribution or warehouse requirements

                                                      6) Spares or service parts requirements

                                                      7) Inter-plant requirements

                                                      8) Samples, demos, trials, etc.


Excluding Demand Sources


Another example of why demand variability can be out of control is when some demand sources are intentionally or unintentionally left out of the equation. Too often, our equation look like this:


Text Box: Supply			=	Forecast

1) People					 
2) Equipment					 
3) Buildings					 
4) Material					 
5) Supplier Capacity












Only the forecast is viewed as a demand source because the other components have been excluded. As a result, the problem of managing demand variability is now magnified. The other sources of demand are no longer part of the equation and, consequently, are not available to absorb any forecast variability. Any variability between actual customer orders and forecasts must be handled by the supply side.


Generally, this approach hasn’t worked very well. Manufacturing and Sales begin erecting fortifications and gun works. We must be brutally honest in identifying all demand sources, despite how we’ve traditionally run the business.


Measuring and Managing Variability

If we knew exactly what the customer wanted to buy for the next six months, what would you produce? Simple, isn’t it. You’d produce exactly what the customer wants. You could also negotiate with suppliers for long-term agreements, work to minimize inventory, level out the people requirements--all actions to reduce the cost of the supply variability.


Reducing variability is a key Total Quality Management issue, as it should be. On the supply side, the reduction of variability is key to improving how well the supply side is performing. How might we work to reduce that variability?


Let’s look at the supply and demand performance measurements. They should indicate the critical elements in the processes.


Supply Performance Measurements

The quality of the supply process is typically measured by minimizing labor and material variances and low unit costs. Bonuses and promotions are awarded with lower unit costs and favorable variances.


High variability causes lost of changes to plans and schedules. These changes cost money and therefore lead to higher costs and unfavorable variances. Therefore, to perform well, based on typical measurements, manufacturing strives for little or no changes to plans and schedules which would reduce process variability on the supply side of the equation.
















Consequently, everybody on the supply side wants a rock-solid plan frozen for the entire planning horizon, which facilitates reducing costs and creating favorable variances. Ideally, the supply plan will have little or no variability if the supply side is to perform well to it’s cost measurements. Notice that we measure and mange the supply side totally independent of the demand side, as if there was no connection at all.


Now let’s look at what’s happening on the other side of the “equals” symbol.


Demand Performance Measurements


            In many companies, some variability of demand is inevitable. How much variability to expect is an important, but difficult, question to answer. Usually, it’s a question that’s ignored. Demand variability is simply not measured, and little effort is put into finding the root causes of the variability and minimizing it. Although a lot of opinions exist, facts are hard to come by.


How are people on the demand side measured? Orders. The more you get the better you are. Some companies even give bonuses for exceeding forecasts. So what does any normal, well balanced person do? Forecast low! Well, if they get more orders and exceed the forecast, their measurements say they’re doing a good job. There is a reward for large variability. There’s also no incentive to encourage reducing variability. The “quality” of the demand planning process doesn’t consider the size of variation or conformance to plan.


Again, notice the lack of congruence between the measurements of supply and demand. The results of this are forecasts that be wrong. Add to this the inherent difficulty of predicting the future, and it’s easy to see why most forecast have lots of variability.


Suppose we decide to measure the variability, and we discover that the forecast variability is ±10% on a monthly basis as shown on the next graph.
















When we overlay this chart with the earlier chart, there’s a discrepancy.

















Now let’s look at a specific example. The supply side is ready to produce 100 units at a low cost (no variability expected or wanted). Demand (Sales) is prepared to sell between 90 and 110 units (100 ±10%). What happens to supply when the orders come it at 90 units? 110 units?


Supply is being rewarded for excellent performance by producing 100 units. Demand is being rewarded for anywhere between 90 and 110 units. If the number of units sold s 90, who has a problem? If you said the company, you’re right. What do we do with the other 10 units? If we sell 110 units, where do the other 10 come from?


The supply side looks at the equation and cries foul. Someone has to accommodate the added the added variability. To clean up. And that someone is the supply side, who see their bonuses and raises go up in smoke. Customer orders are being promised in less than the agreed upon lead-time. Some weeks order volume is up; other weeks it’s down. One week there’s overtime; the next week the machines are idle for a day or two. No one knows what to expect--damn the forecast and the people who made it! Meanwhile, on the demand side, no one understands why the supply side can’t ship customer orders on time after everyone has knocked themselves out to get them.


Managing the Whole Equation

What went wrong?


We got trapped by the silo effect and ignored the process of balancing the equation. We created the silos, the vertical structures within our companies, to ease our growth traumas. Engineering, Purchasing, Finance and Production are examples of these vertical structures. But the silos have blinded us to the fact that most of our critical business processes are horizontal, cutting across departmental lines. The process of balancing supply and demand is certainly one of these. The result is that we have made supply decisions separate from demand decisions.


Manufacturing companies can no longer afford to isolate supply from demand or demand from supply. Sales forecasting should be viewed as one part of a company-wide process for managing demand and irrevocably tied to the sister process of managing supply. We’ve got to change both the existing practices and the measurements to allow us to focus on managing the both sides of the equation. The solution is to change the formula so that all  demand and supply sources are considered in managing the business. This allows increasing or reducing the size of the customer order backlog or finished goods inventory to act as a buffer to some of the forecast variability. The remainder of the variability must be absorbed by the supply side of the equation.


Determining this portion of variability becomes a joint decision on both sides of the equation. Typically, this is a monthly decision. Some months, the customer order backlog can’t be expanded, and accordingly, finished goods inventory may decrease. Supply must work overtime and the suppliers are asked to use their flexibility. Other months, customer orders aren’t coming in as planned and the backlog is reduced.


As a result, supply performance is no longer measured only in unit costs and variances, but also on absorbing the agreed-upon demand variability. The quality of the supply and demand processes can now be measured by how well they meet their internal customers’ expectations. Supply can now be rewarded for maximizing flexibility within cost constraints. Demand is now measured on hitting the agreed-upon target within the accepted variability. Now the equation balances. Supply equals demand.


Action Items


Here are some action items to improve the supply/demand processes in your business.


1) Identify the specific individual components of the supply and demand portions of your business.

2) Establish an unbiased performance measurement for each demand component.

            3) Establish an unbiased performance measurement for each supply element.

            4) Agree on the expected (planned) variability of each supply and demand element.

5) Document the process for evaluating, measuring, and reconciling the variability between supply and demand.

             6) Remove obstacles to communication between areas or departments.