Confidence is not
a forecast

Background and Methodology

1. The Study

Between 2001 and 2011, researchers surveyed chief financial officers at major US companies every quarter. Each was asked for a range they were 80 percent confident would contain the return of the US stock market over the following year. Over ten years the study collected more than 13,300 such ranges from senior, numerate, professionally careful executives. If the executives were as well calibrated as they believed, roughly 80 percent of the ranges would have contained the actual return.

2. The Key Findings

• 36% Hit Rate: The actual return fell inside the stated ranges 36 percent of the time, against an expected 80 percent.
• Confidence Without Calibration: The executives' confidence was sincere. Their ability to bound uncertainty did not match it, and the pattern held across the full decade, indicating that experience alone did not correct it.

The Planning Fallacy

A related body of research examines how people estimate the time their own work will take. The findings are consistent: people complete tasks later than they predict, including tasks they have performed many times before. The optimism persists even when participants are explicitly instructed to assume things will go wrong and to pad their estimates accordingly.

Why Revenue Targets Inherit Both Errors

A revenue target set from the top and defended on confidence combines the two findings above. It carries overconfidence about the range of possible outcomes, and it carries optimism about how long the underlying work will take. Because neither error is visible from inside the target, the number typically survives every quarterly review until the quarter in which it fails.

A Calculated Alternative

The alternative replaces conviction with a chain of counts that can be verified as the year proceeds. The revenue target is divided by the average deal size to give the number of deals required. That figure is divided by the actual proposal close rate, rather than the rate used in board presentations, to give the number of proposals. Proposals are divided by the share of first conversations that lead to a proposal, and conversations are divided by the share of qualified inquiries that become a conversation, using a written definition of “qualified” that two people would apply identically without conferring.

Worked Example

With a target of $1,200,000 and an average deal of $60,000, the target requires 20 deals. If one proposal in four closes, 80 proposals are needed. If half of first conversations produce a proposal, 160 conversations are needed, or about three per week. If four qualified inquiries in ten become a conversation, the target requires 400 qualified inquiries, or roughly eight per week. The figures are illustrative and should be replaced with the business’s own historical rates.

Implications for Your Business

• Weekly Verification: Eight qualified inquiries per week can be counted every Friday. A $1,200,000 annual figure cannot be checked against anything until it is too late to act.
• Early Detection: If the weekly conversation count exceeds the capacity of the available staff, the calculation exposes a staffing constraint months before it would otherwise appear as a missed year.

In Conclusion

The miscalibration research shows that confidence, even among highly qualified professionals, is a poor proxy for accuracy. A target built as a chain of weekly counts is not more optimistic or more conservative than a target defended on conviction. It is simply checkable, which is the property a forecast requires in order to be useful.