Forecasting

In a webcast entitled Worst Practices for Forecasting, SAS Marketing Manager for Forecasting Mike Gilliland points out some of the major pitfalls in business forecasting. One of the biggest problems, he says, is that companies often invest too much time and too much money but generate little return. Why?

Hes glad you asked.

Awareness of some of the worst practices of forecasting can go a long way toward making forecasting more efficient and more effective. Its not always appropriate to blame the forecast if the numbers dont match up its also important to look at how those numbers were produced.

The politics of forecasting
One problem companies often fall prey to is failing to understand the politics that can impact forecasting. Everyone involved in the forecasting process has agendas, biases and motivations, says Gilliland, and these can affect results.

A salesperson may forecast a low demand for widgets because this sets a low, easily achieved goal. That same salesperson may later forecast a high demand, this time to ensure sufficient inventory. Equally, the person who invented a new product or introduced a new service is unlikely to predict that that product or service is going to be a bust, even if all indicators point toward failure.

When it comes down to the question of have enough on hand to ensure good service or dont have too much to save operating capital, everyone has an opinion influenced by their own needs, whether they realize it or not. And those opinions can alter forecasts.

Says Gilliland, a forecast should be an unbiased, best guess. There are lots of numbers floating around companies: targets, sales quotas, financial budgets. All of these numbers have their place and purpose, says Gilliland, but they should not be confused with forecasts. Too often, excessive optimism about future demand takes the place of sober rationality. Its good to have goals that stretch your team, says Gilliland, but its important that forecasts be based on the realities of the marketplace. In other words, the forecast should be your best guess of whats really going to happen, not a reflection of your hopes.

According to Gilliland, companies often put too much faith in Unconstrained True Demand. UTD is a measure of what customers want and when they want it, a measure which, Gilliland says, is nebulous. How can you truly know the demand for your product when that product sold out? Could you have sold twice as many? Three times? Or had the demand peaked just at the point at which you ran short of supply? If you supply a service like a phone-in help line, what if youre understaffed? How many more people might you have served with adequate staff? How many staff did you actually need? With so many possible variables, UTD can never be truly accurate. And yet, companies frequently rely on it when making forecasts.

Says Gilliland, a forecast should represent the voice of the marketplace customer demand. However, this should be tempered by your companys ability to meet that demand.

Mind the gap
When new forecasts are done, there is often a gap between the new forecast and the old plan that was based on previous forecasts. The tendency is to blame the old forecast for producing such inaccurate numbers. Instead, says Gilliland, we should see that gap as a blessing. If you sold fewer of your product than was forecast, consider lowering the price or increasing your marketing or reducing your production. If you sold out, it might be time to think about increasing the price, reducing marketing activities or upping production. Best practice says to use this as an opportunity to make changes that align reality with the forecasting numbers.

Worst practice is to change your new forecast to match your existing plans without bothering to find ways to close the gap.

Blaming the messenger
Companies that place too much blame on the forecast may be missing some other important factors that led to unforeseen results. Forecasts, no matter how sophisticated the metrics, cant possibly predict every variable of the marketplace. And sometimes, says Gilliland, the unexpected comes from within.

Make the demand foreseeable, Gilliland says. There are practices among sales forces that have significant impacts on demand: promotional events, price changes, even a sales contest or an end-of-quarter push to make targets can create volatility. The more stable, the smoother the demand, the fewer spikes and troughs, the easier it becomes to forecast the future.

Any knucklehead can forecast a straight line, says Gilliland. Increased demand volatility leads to reduced forecast accuracy. While it may look good to have a record week in sales, such extremes in consumer demand may ultimately make forecasting less effective.

Measured success
Most companies focus on the accuracy of a forecast: how close was the guess to the reality? Yet, says Gilliland, accuracy may not be the best or only measure of a tools value.

FVA, or forecast value added, is a measure of the change in a forecasting performance metric (such as accuracy, Mean Absolute Percent Error or bias) that can be attributed to a particular step or participant in the forecasting process. The results of making a forecasting step are compared to results that would have occurred, had the forecasting step not been taken. A positive FVA indicates that the forecasting activity is worthwhile and makes the forecast better. A negative FVA indicates that the activity actually makes the forecast worse. How much accuracy can you reasonably expect, given your demand patterns, product history, etc? FVA can tell you if your tools are efficient and effective; this is far more useful than simply knowing how far off your forecasts were.

The idea, according to Gilliland, is to fill your toolbox with forecasting tools that are truly effective. Poor tools or good tools poorly used not only waste money and time, they can also skew results.

You can hear Mike Gillilands webcast in its entirety on the SAS website: www.sas.com.

Author: Shannon Perry, journalist, theforecaster