The art, science, and frustration of sales forecasting

Yogi Berra said it best: It’s tough to make predictions, especially about the future

The FABRICATOR June 2013
June 28, 2013
By: Dick Kallage

Forecasting, while seemingly futile at times, is something fabricators should take seriously instead of dismissing it as some hocus-pocus best left to Wall Street analysts, politicians, and tarot card readers.

At The FABRICATOR®’s Leadership Summit held in late February in Tampa was a packed session on forecasting. I was able to contribute some baseline thoughts and principles on the subject for the session’s able moderator, Lori Tapani, co-president of Wyoming Machine, a Minnesota fabricator. The session was very interactive and lively, with inputs and discussions among a range of company executives, from major OEMs to smaller specialty fabricators. What I got out of this was the impression that our community agrees on little about the subject of forecasting beyond the fact that Yogi, as usual, was right. In fact, more than one attendee thought that the entire exercise of forecasting was, in fact, futile.

I can certainly empathize, but I have to disagree on the notion of futility. Forecasting is not futile. It’s just a messy process at times and by its very nature inaccurate. The issue that leads to the notion of futility is always one of accuracy. If the errors are almost always very large, then the process of forecasting would indeed seem rather meaningless.

However, before you stop reading, let’s examine a few things about forecasting that I think you will find interesting and relevant. First, let’s put a frame around the topic. We’re talking about forecasting future performance for custom product producers, including fabricators and specialty manufacturers. In particular, we will focus on the most difficult thing to forecast: future revenue levels.

The questions we’ll ask are:

  • Is forecasting even necessary?
  • If so, is it possible to derive results that are meaningful?
  • If it’s necessary and results can be meaningful, what do we need to know to construct a functional forecasting process?

Let’s see if I can convince you that forecasting is something you should take seriously instead of dismissing it as some hocus-pocus best left to Wall Street analysts, politicians, and tarot card readers.

Why Should You Bother?

We need to estimate the future to define what resources we are going to need and what’s available. Forecasting obviously is essential for determining future cash flows, staffing needs, ability to meet loan covenants, and a host of other items critical to operating a business.

Forecasting is at the heart of improvement, because the improvement choices we make are tightly coupled to the demand levels we predict. If sales are forecast to be flat or even down in the next, say, 12 months, we will elect to execute improvement actions that are different, and on a different scale, than if we were forecasting a 20 percent increase.

So it’s pretty obvious that we need to forecast the future, even if the process is imperfect, because we simply cannot not forecast.

Why Forecasting Is Meaningful

Before globalization, forecasting was easier. Most OEMs provided relatively firm forecasts to their supply base and actually stuck to them, usually by contract. So, short-term revenue forecasts were relatively easy, though even then they weren’t totally accurate. (A perfectly accurate forecast isn’t really possible in a free market.)

But during the past 10 years, I have noticed more OEMs moving away from forecasting for their supply base, and most of those that still do supply only guideline information—nothing firm. This has raised the question of meaningfulness. If our customers don’t have a solid gauge of future demand, then how can we? Well, unless the OEM is very poorly managed, it does have a firm idea of future demand, and a formal process for assigning numbers to it. What OEMs don’t want to do is share it with suppliers as part of a contractual arrangement and be held liable for parts or assemblies that they have to pay for, but in fact didn’t need due to inevitable errors in their forecast.

But forecasting is still very meaningful, even if we don’t get information from customers. The process itself, despite its inherent inaccuracies, forces us to examine, to the best of our abilities, the dynamics of our markets and, in particular, our largest customers. This is a fundamental business process. Companies that don’t do this tend to get whipped around unnecessarily severely.

A far better way is to gather all of the data, impressions, educated guesses, and all other pertinent inputs; put them into some sort of process; and crunch an output that quantifies (in dollars) the demand for some future period. This process, done regularly, provides a wealth of critical information that is a source of competitive advantage.

So are the results from forecasting meaningful? Absolutely. If we have a process, take it seriously, and continually improve it.

The Forecasting Process

Restricting ourselves to forecasting future revenue, we can build a process that works by first understanding some fundamentals about this type of forecast. A forecast has two components: how much and when. The most common when component is forecast shipments for specific months.

The forecast is a function of a large number of variables, most of which we can’t influence. Globalization has increased the number of these variables, as has the trend toward low inventory channels throughout entire supply systems. The number of variables is directly related to the forecast’s error potential. But we can focus on a relatively small number of variables and still have a functional, worthwhile forecasting system.

To forecast revenue, we really must forecast bookings. We age those bookings to when the resulting shipments occur, and add them to existing backlog. What follows are some tips and techniques that apply in creating a forecasting process. We can know a lot more than we think we can. There is an amazing amount of information available that helps us forecast to a reasonable degree of accuracy.

  • For most job shops, many statistical techniques, including linear regression based on history, are not strictly valid (too few customers make up the preponderance of sales), but averages of bookings over, say, the prior six months are a good place to start.
  • For the large number of customers that usually generate 20 to 30 percent of sales, you can use averages or linear regression of their combined historical bookings, if you have more than 20 customers. If one or more of these customers is about to become large, treat them as one of the few large customers, as described next.
  • For the few customers that typically generate 70 to 80 percent of sales, treat each one individually, starting with average monthly bookings and adding or subtracting changes from product launches or discontinuations. For these customers, you must have an intelligence system that captures all available information regarding their sales, inventory levels, and production rates. There are many sources of this information, from the 10Q reports of public companies to contacts in production, sales, and engineering. If you do this well, you will rarely get blindsided.
  • Track bookings by customer for a long period of time. You will be able to capture subtle cycles and other rhythms by studying the bookings’ trajectories.
  • Short-term forecasting is often more inaccurate than long-term forecasting. This is because slight changes in when those bookings occur can cause large errors in how much (that is, quantities) the forecast predicts for a given slice of time. It’s annoying, but shifts of a week or even a month aren’t the end of the world. Focus on reducing the how much error.
  • To modulate expectations, look closely at the larger economy and its indicators. Among the leading indicators that I use are (outlook validity in parentheses):
  • Raw materials (steel) production and transportation indices (6 months).
  • Export data, especially to Europe, South America (3 months).
  • Institute for Supply Manage-ment™’s Purchasing Managers Index (PMI) (1-3 months).
  • Credit data (1-3 months).
  • Key customer and customers of key customers’ financial data and comments (3-6 months).
  • Industrial inventory to sales data (3-6 months).
  • Estimates of capital equipment expenditures: direction up/down (3 months).
  • Consensus GDP forecasts: This puts a baseline on growth based on current business.
This sounds a bit complicated and messy, but once you’ve got your system in place, it’s actually pretty straightforward. So keep at it, and keep improving. It is totally worth the effort.
Dick Kallage was a management consultant to the metal fabricating industry. Kallage was the author of The FABRICATOR's "Improvement Insights" column from May 2012 to March 2016.

Published In...



The FABRICATOR is North America's leading magazine for the metal forming and fabricating industry. The magazine delivers the news, technical articles, and case histories that enable fabricators to do their jobs more efficiently. The FABRICATOR has served the industry since 1971.

Preview the Digital Edition

Subscribe to The FABRICATOR

Read more from this issue

Related Companies