I was once an analyst for an aircraft component supplier. The Division Vice-President asked me for an estimate of the market size and of our sales for the next year. I had access to the best data in the whole world. My data listed every commercial airliner in the world (over 40,000 of them), by model and type, by owner, by tailnumber. It included hours flown for each aircraft, by month. I could tie this data to our records for each customer and tail number. I built a huge spreadsheet model and pasted the spreadsheets all over my office wall.

Then I went to the VP and said, “The market will grow 3% next year.”

The VP said, “That’s not good enough. I need 5% to tell the President.”

So I said, “I will reevaluate the data.”

I knew that despite all the data, there were still unknowns that had to be estimated. Ninety-two of them. How many hours would each airline fly each aircraft next year? What heads-of-state or corporate customers would upgrade this year? Which planes would be retired. How many aircraft of each type would each manufacturer sell? et cetera. Each of those estimates had a probable range.

So, I opened the gigantic model. Set a goal-seek to set the market to 5% adjusting ONE variable. The resulting new assumption was still inside the probable range. I told the new result to the VP and he was happier than a pig in mud.

Was I being dishonest?

No, and yes. No: The changed variable was just as valid as the original assumption. The math logic was perfect. Yes: I could have assumed the worst possible case for each variable, within probable limits. The forecast would have been negative. Alternatively, I could have assumed the best possible case. The growth would have been astronomical. In the aggregate, both outcomes would be highly improbable.

What are the lessons here?

1. Look for hidden assumptions. Don’t just look at the math and certainly don’t accept the assertion “because the computer said…”

2. Question the values of the assumptions. Look for bias. Do a sensitivity analysis of the most salient assumptions.

3. Remember the First Rule of the MBA Analyst:
Step 1: Gather all the data you can.
Step 2: Find out the answer your boss expects.
Step 3: Manipulate Step 1 to Match Step 2. Then make sure the output is obviously from a computer.