During these turbulent and unpredictable times, forecasting is less fun than ever! But forecast we must, if we are to manage our companies competitively and well. The question is, “how to forecast as best we can?”.
Back in 2002, I co-authored with Tom Wallace our first of six books, Sales Forecasting-A New Approach. That book contributed to the progression of our current day practices. At that time, the standard practice was to forecast the future in a highly granular fashion (SKU by customer) for the full S&OP horizon. In the book, we called this approach “forecasting in the suicide quadrant” (Chapter #9). We called it that because it had two serious flaws:
- The detail produced a fog of complexity where we could’t see the forest for the trees.
- It was totally customer-centric, failing to see or observe any market trends and patterns.
Shortly thereafter, forecasters began to add up the detail and look at the bigger – the volumetric picture organized by product or production family. That represented a real improvement but it was still a customer-centric forecast, not reflecting market tends and patterns.
Today, best practice for volumetric forecasting projects the future by market-facing families, with multiple views. This results in a Consensus Forecast that predicts the volumetric future with reasoned, reasonable, credible, and transparent logic. That forecast is then translated to production-family volumes through the use of simplifying assumptions about mix.
The Principle of Multiple View Forecasting
In this piece, I’d like to focus on understanding the principle of “multiple view” forecasting. Remember the story of three blind people exploring an elephant – one discovered a rope (the tail), one a hose (the trunk), and one a wall (the side). It was not until they shared and compared their views that the picture of an elephant became apparent – no single view was sufficient.
Forecasting is much the same way. People in different functions within an organization have different views about the future – they see things through different prisms. Each is credible in its own way, but no one view tells the whole story. We might describe these multiple views as follows:
1. Market View – this is expressed by Market-Facing families (for an example, see insert), correlated to total markets based on leading market indicators. Total market data and predictions are often gotten from outside third parties. The company’s market-based forecast is then derived from mathematical regression and/or anecdotal correlations to this market forecast. It serves to look beyond the customers to the markets they serve, and is thereby not limited by the customers’ knowledge, bias, or shortcomings. This market perspective is used as one input to the company’s volumetric forecast.
2. Customer View – typically expressed in SKU’s, end items, or orders by customer indicating what each customer will need for the next three to six months. In some cases it’s the customer’s schedule (make-to-order), and in others it’s the company’s forecast (make-to-stock) of what they think their customers will need – or some combination of the two. This information is used primarily as an input to the Master Scheduling practice. Aggregated into the Market-Facing Families, it is a second input to developing the company’s volumetric forecast.
3. Historical View – this projects the company’s history into the future by a mathematical model. This modeling projects the aggregate Market-Facing Families for the volumetric planning horizon. A technique called “Focus Forecasting” (fitted forecast) is typically used. This modeled future reflects the past based on statistical trends and patterns that can be observed and modeled. It is a third input to the company’s volumetric forecast.
Note: There may be other “views” that come to into play such as brand view and channel view etc.
Reconciling These Various Views
Each of these views could be quite different, each having individual credibility, but based on different perspectives. Sometimes one or more of these may have greater credibility or weight than another. Reconciling them into a single collaborative view is a human activity that is necessary to get all functional departments operating from the same set of demand numbers. This has become a best-in-class volumetric forecasting practice. The objective is to develop a Consensus Forecast that is reasoned, reasonable, credible and fully transparent, that everyone buys in to.
Step #2 of the monthly executive S&OP process is known as Demand Planning, where volumetric forecasting by Market-Facing Family is performed. Reconciliation of the multiple views into a Consensus Forecast is the final activity of this step.
This reconciliation activity is sometimes called the Demand Agreement Meeting (or the DAM Meeting for short?). Although conceptually simple, this is not an easy meeting to run, because it brings into sharp focus the potentially divergent views of sales (the customer view), marketing (the market view), new product, supply chain, and finance, etc. into sharp contrast. Attending this meeting are executive representatives from each discipline within the business.
Supported by data, the objective of this meeting is to expose disagreement without being disagreeable, arrive at one set of volumetric demand numbers by which to run the business. This becomes the Consensus Forecast.
Applying “Puts & Takes”
Once reconciliation has been achieved, there is one thing left to do: Compare the Consensus Forecast (converted to revenue) to the “last call”. If the Consensus Forecast compares unfavorably to the last call, appropriate actions are then identified to bring the forecast or the last call to where it needs to be. These actions are documented, with progress being tracked in future months.
The best view of the future comes from blending multiple views into one set of demand numbers, called the Consensus Forecast. This open disagreement and reconciliation should happen routinely as part of the executive S&OP process, and has become best practice for volumetric forecasting. In the words of one S&OP Manager, “ …because each input has strengths and weaknesses, having all views brought to the table brings about more educated and better forecasts. Typically . . . the more initial disagreement, the better the forecast.”