Newsletter Sign-up

"Welcome back. I need your forecast for 2011!" Sound familiar? I hope that everyone had wonderful summer vacations and is ready to dive full steam into forecasting and budgeting season for 2011.

If not, in last month's column we tried to help you get ready by discussing the three major issues with the way many organizations forecast today. And today, we will explore the specific techniques for getting organizations comfortable with dealing with key variables that have some degree of uncertainty and are outside of direct control. We will work through a step-by-step approach for collecting a forecast as a distribution, instead of as a single number. Finally, we will see how this approach helps us calculate our perceived odds of success.

So how do you get a distribution? One approach is to follow the steps below for each variable you want expert opinion on. The variable could be anything from total 2011 revenue to 2011 customer demand for widget X, to the Fed funds target rate at Dec. 31, 2011.

Step 1: Consistently identify, define and articulate the variable (e.g., "We would like you opinion on the what the Fed funds target rate at Dec. 31, 2011"). In this example, this focus will help you avoid the situation where one person answers the Fed funds rate question and another tells you how many times they think the FOMC will change rates between January and December.

Step 1b (OPTIONAL; USE CAUTION): Provide the expert any "data" that you want them to consider before providing their estimates. Say, a Bloomberg chart showing the market's view on this variable before asking them about the Fed funds target rate. Understand that any data that you show them immediately before the assessment will both inform and likely bias the estimates they provide you. So if your data is questionable, I would not show it.

Step 2: Try to remove some anchoring bias by asking for the extremes. To continue with our example, you could ask:

-- What is the lowest number you possibly could imagine for the Fed funds target rate on Dec. 31, 2011? Please give me the number where you feel that there is only a 1 percent chance of the actual value being any lower.

-- What is the highest number you possibly could imagine for the Fed funds target rate on Dec. 31, 2011? Please give me the number where you feel that there is only a 1 percent chance of the actual value being any higher.

Step 3: Ask for "low", "high" and "base" case estimates. In other words:

-- What would you consider a reasonable number to use for the "low case"? Please give me the number where you feel that there is a 90 percent chance of the actual value being higher than this number.

-- What would you consider a reasonable number to use for the "high case"? Please give me the number where you feel that there is a 10 percent chance of the actual value being higher than this number.

-- What would you consider a reasonable number to use for the "base case"? Please give me the number where you feel that there is a 50-50 chance of the actual value being higher/lower than this number.

Step 4: Check the values: If the expert is not indifferent to each of these bets, adjust the assessment numbers until test are passed. So ask:

-- Would you rather bet on the actual value being below (the value given in the first question in step 3) or above (the value given in the second question in Step 3)?

-- Would you rather bet on the actual value being between (the value given in the first question in step 3) and (the value given in the third question in Step 3) or between (the value given in the third question in Step 3) and (the value given in the second question in Step 3)

-- Would you rather bet on the actual value being above or below (the value given in the third question in Step 3)?

Step 5: Ask the expert what would have to occur for the actual value to end up in different portions of the distribution.

-- In the lower quartile (First to 25th percentile). Ask: Imagine that we are now at Dec. 31, 2011, and the Fed funds target rate is (insert answer to low case/10th percentile). What could have transpired in order for this to be the case?"

Listen to the expert explain the different things that could have transpired. Take good notes and watch for them sighting values for key variables (e.g., worsening unemployment, or the U.S. economy still in a prolonged recession).

-- In the upper quartile (76th to 99th percentile). Ask: Imagine that we are now at Dec. 31, 2011, and the Fed funds target rate is (insert answer to high case/90th percentile). What could have transpired in order for this to be the case?

Listen to the expert explain the different things that could have transpired. Take good notes and watch for them sighting values for key variables (e.g., improvements in employment or the U.S. economy recovers quickly).

-- In the middle quartile (26th to 75th percentile). Ask: Imagine that we are now at Dec. 31, 2011, and the Fed funds target rate is (insert answer to base case/50th percentile). What could have transpired in order for this to be the case?

Listen to the expert explain the different things that could have transpired. Take good notes and watch for them sighting values for key variables (e.g., steady unemployment or the U.S. economy stabilizes and shows signs of recovery).

-- In the extremes. Ask: Imagine that we are now at Dec. 31, 2011, and the Fed funds target rate is above the figure you gave as the highest (or 99th percentile). What could have transpired in order for this to be the case?

Listen to the expert explain the different things that could have transpired. Take good notes and watch for them sighting values for key variables (e.g.. huge employment improvements, the U.S. economy roars back to life stronger than ever, real estate values rise to historical highs, all the money Bernie Madoff stole from people is found in a bank vault and all investors are made whole).

Record the data so that, one, you can easily confirm it with the expert and, two, compare or combine it with other expert input to use for analysis.

Sure, this is a lot more detailed than asking for a baseline, low and high. However, there is value in the details. For whatever variable you assessed, you now have the perceived odds for any given value in your distribution. For example, if the variable you assessed is the 2011 revenue for widget X" and the value recorded at the 50th percentile on your distribution is $10 million, then you now know that the perceived odds say that there is a 50 percent chance that widget X will make $10 million or greater in revenue for you in 2011. You can then do this across your entire distribution.

Congratulations. You are now a bookie!

DAVID M. WONG is director of enterprise risk management at CME Group, the world's largest and most diverse derivatives exchange.

(Editor's note: Read last month's column, "Forecast Says 50 Degrees for Next Year: Part 1," here.)

September 21, 2010

Copyright 2010© LRP Publications