The Forecast isn’t the Forecast, Until it’s YOUR Forecast!

Wouldn’t it be nice to have a 100% accurate forecast? For sure. Of course, we all know that this isn’t possible, but to quote Henri Poincare in The Foundations of Science:

"It is far better to foresee even without certainty than not to foresee at all."

The ultimate goal of demand forecasting is to accurately predict what will happen at a defined time in the future.  For many of you, your only evaluations of accuracy are:

  • Your forecast accuracy (or error) metric
  • Your demand exception count

So many of the people we work with dread demand exceptions. Their primary goal is to have ZERO demand exceptions. For years, you (and we) have focused on demand exceptions to help make your forecast more accurate. It seems to make sense, but in a way, it’s too little too late, isn’t it?

By the time you’re evaluating exceptions, demand has occurred and the facts are the facts. What if there was a way to improve the forecast before demand occurs, and at the same time eliminate the process of trying to discover why an exception happened in the first place?

The first step is to think about forecasting differently

You might want to read this twice…or more:

The forecast is NOT limited to what your system’s statistical forecast says.

It is ALL of the demand that will occur in the future, regardless of the cause.

If you’ve worked with us for a long time, this could be a revelation.  It was for many of us.

Consider this, whether it’s “regular” demand, or promotional demand, or committed orders, or whatever causes the customer to buy from you…it’s demand. And therefore, it should be part of your forecast. The only reason it may not be in the forecast today is a limitation of your system. Think of a forecast as an accumulation of all the reasons a customer buys from you.  See the example below:


How can you improve your forecast and reduce demand exceptions?

Turn the forecast process on it’s head…get in front of demand.  The more information you can provide to your forecast prior to the demand occurring, the more accurate the forecast will be

Pre-emptively include the many factors that will influence demand in the forecast:

  • Temporary causal impacts, such as weather, advertising, sales drives, etc.
  • Permanent base rate impacts, like competition, cost changes, new customers, etc.
  • Incremental known impacts, like bookings, scheduled orders, pre-orders, etc.

Why has this been happening to me?

Most systems rely solely on the statistical forecast to predict demand. The statistical (stochastic, random) segment of the forecast represents the units that on average (give or take measured deviation) are expected to sell over a defined period of time, based on demand that occurred in the past.

What the statistical forecast doesn’t reflect is demand that hasn’t occurred in the past, but is known or expected to occur in the future. That’s potentially a big gap! Especially if your market is changing in any way.

I’ve always said, “If you know something that the system doesn’t, then let it know.” The forecast is only as good as the information fed to it. So, if you (or someone else) know something more than what the historical data represents, you can let your system know before demand occurs in a number of ways (assuming you have Blue Ridge):

  • Have the solution calculate causal forecasts (event, promo, etc.) using the Event Forecasting application
  • Enter your estimate of the causal forecast into the Event function
  • Collaborate with colleagues, suppliers, even customers in the Aggregate Forecast Management application, and augment or replace the statistical forecast with your forecast decision.

This way the system can properly accumulate the forecast, and account for demand before it occurs. These insights can be enormous in improving your forecast accuracy.

C’mon!  Can You Really Do That?

To make the forecast more accurate, we’ve developed the ability to create segments of the forecast that augment the statistical forecast.

  1. The Blue Ridge Event Forecasting application captures the influence of past events so that it can create a forecast for future events with similar characteristics. Event Forecasting can create event forecasts for SKU/locations and even specific customers, by capturing how certain causal factors affect demand temporarily.
    Since the causal forecast segment is part of the total forecast, replenishment uses that forecast to automatically adjust inventories to support the temporary demand, and return to its “base” level after the event.
  2. Dependent Demand Forecasting represents demand that is highly expected or even pre-determined, such as order schedules, committed or booked orders, known as Deterministic Demand. This effect can be interfaced or be entered directly in to SCP and becomes another segment of the forecast.
  3. Aggregate Forecast Management (AFM) enables people to have input into the forecast and to accumulate data in many ways to determine what makes sense to determine the future. This tool allows supply chain, merchandising, sales, suppliers and even customers to have input to the forecast in the effort to more accurately represent future demand.
    I’ve always found that a good line of communication with Sales and Marketing is invaluable. A large promotion or new customer demand, if planned in advance, can be included in the forecast. Instead of being blindsided, you will improve the forecast and the results, and gain the confidence of your team by providing better fulfillment.

With these inputs of forecast information, your forecast will reflect the proper demand, fulfillment will be more effective, and (last, but not least!) an exception can be eliminated, so you don’t have to scramble around trying to figure out what happened, then determine whether to adjust forecast and clean up history after the fact.

Think of the forecast as a series of segments.  With those segments represented, the total forecast will represent what history tells you and what the future tells you. Your forecasts will be more accurate and your workload will be less, and you’ll pre-empt all the other downfalls of incomplete forecasts.

And remember…Forecasting is a GREAT profession!

Jim Klekamp

Senior Business Consultant