Main content

John R. Dougherty

The Problem

Having trouble forecasting isn’t something new.  In fact, it’s the single most common problem shared by every company, large or small, in manufacturing and distribution.  Though the problem is not new, it often worsens as time goes by.  Why—because of the increasing volatility of markets and the shortening life cycle of products.  In addition, as competition grows more aggressive and customers more demanding, the potential impact of inaccurate forecasts becomes more significant. 

Naturally “lumpy” demand patterns: Unfortunately, customers don’t buy the same amount of every product they need, every day.  Often they don’t even do it every week or every month.  Why—inconvenience, cost, and irregular demand patterns from their customers or the ultimate consumer.  Ordering, transportation, receiving, handling, set-up and manufacturing, and other related costs trigger lot sizing decisions at every link in the supply chain.  For example, though the ultimate consumer of a pharmaceutical may consume in an even pattern, say one or two tablets a day, there is much lot sizing between them and the manufacturer.  The consumer may buy a 30, 60 or even a 90-day supply at a time, based on a doctor’s prescription, medical insurance rules, deductibles and prices, etc.  Likewise, the pharmacy, distributors, and ultimate manufacturers of the pharmaceutical may also replenish their supplies in inconsistent quantities based on varying demand patterns, marketing promotions, pricing protocols, cash flow considerations, and the response to sales targets and agreements.

Two Kinds of Inaccuracy

 A forecast error or deviation is defined as the difference between the demand that was forecast and the actual demand that occurred.  It is generally measured once a month by family or by product.  Sometimes quarterly or year-to-date totals are also reviewed.  With the proliferation of markets, products, and customer-sensitive options, it becomes very difficult to analyze all these forecasting errors in companies with growing product lines.  So, in the interest of time and expedience, often the only errors that are analyzed are the most significant variances that occurred, last month.  This is far too narrow a perspective. 

There are two significant but very different contributors to any forecast error.  The first is called bias.  Bias represents a consistent forecast error in the same direction (actual sales usually being above forecast or below forecast) over a period of time.  This is the most difficult and harmful form of forecast inaccuracy, because over time either inventories will build or customer service problems will proliferate.  The only way to detect bias is to look at the actual sales vs. the forecast over a longer period of time in an attempt to identify the pattern.  Often bias can be caused by undetected seasonal demand patterns, cycles, or long-term trends.  Sometimes the bias is exaggerated by organizational or interpersonal considerations such as always forecasting low so that sales and marketing can “beat the numbers.”  In other cases forecasting is always high to justify new product introductions, capacity expansion, inventory availability, meeting an aggressive revenue growth target from top management, or to justify an advertising, spending or staffing budget.

It is critical to identify and adjust for these or any other causes of bias, since it will have the largest cumulative impact on costs, inventory, and service.

“Random” or “normal” variation above and below the mean or average demand is the other portion of forecast error.  Once the bias is removed, it means that the forecast is just as likely to be low as it is to be high in any time period.  And, over time, the plusses and minuses will balance out and the total forecast will closely approximate the total actual sales.  This “lack of linear demand pattern” is more prevalent in some products than others.  It is affected by the behaviors, practices, and economics of Demand Chain Partners that affect the timing and quantity of the product being ordered.

For example, low volume, low cost products are much more likely to have non-linear demand since it is easier and more cost efficient for the demand chain customers to lot size or order the materials less often, without paying the price of a large investment in inventory.  Critically needed is the ability to analyze forecast vs. actual sales over longer periods of time to allow the responsible people to differentiate between bias and “normal variation,” and then act appropriately.

      Otherwise, just looking at last month’s sales could trigger adjustments of forecasts up and down in response to normal variations when, in fact, the average forecast is accurate and what is needed is the ability to accommodate normal variations.  Otherwise, undetected bias may not be noticed for several months, until inventories or backorders are excessively high.

So What’s the Answer?

Attack the bias first: Using whatever simple analytical or reporting tools are easily available, begin to analyze forecast errors for bias by looking at the data over a longer period of time.  Then correct the bias; that is, change the forecast to the best estimate of average demand, such that the month-to-month errors will be relatively evenly spread above and below the new forecast.  Often this means getting management’s concurrence to change the numbers to fall below or above current business targets.  This requires a management culture that accepts the fact that the current forecasts are not necessarily the business objective.  It is critical to accept these differences, discuss them, highlight them, and then deal with them. 

Then plan for variation: Even with bias totally eliminated, normal week-to-week or month-to-month demand variation needs to be accommodated.  There are a wide variety of techniques to help accommodate this, including:

  • Moving to finish-to-order or make-to-order production approaches.

  • Shortening supply chain lead times.

  • Increasing manufacturing flexibility.

  • Planning for safety capacity.

  • Material overplanning.

  • Safety stock / inventory.

  • Involve all Supply Chain and Demand Chain Partners.

  • Establish reasonable expectations.

  • Use Sales & Operations Planning and Forecast Accuracy measurements to monitor performance and focus efforts.

Summary

Though there are numerous tool and techniques to improve forecasting, consciously separating bias from “normal” variation helps ensure that the right tools are used for each type of error.

For more information on improving forecasting, customer service, and business effectiveness, call 603 528-0840, e-mail This email address is being protected from spambots. You need JavaScript enabled to view it., or write to 100 Fox Hill Road, Belmont, NH 03220 for copies of the following articles by John Dougherty:

      “Forecasting - Can You Do It Good Enough?”

      “Scheduling to Keep Your Customers Happy”

       “The Role of Sales & Marketing in Planning and Scheduling”

      “Re-engineering Your Business Planning Through Sales & Operations Planning”

_______________________________________________

 

FEEDBACK

If you have specific questions about this article or want to discuss it with the author, call John Dougherty at 1 603 528-0840.

The Partners for Excellence specialize in helping companies set up comprehensive measurement programs and improving overall resource management performance.  Contact us at 1 603 528-0840 or email This email address is being protected from spambots. You need JavaScript enabled to view it..