Overconfidence bias in executive decisions: how a score corrects it

What is overconfidence bias and how does it appear in executive decisions?

Overconfidence bias is the habit of trusting one’s judgment more than its track record supports. Wikipedia lists 3 forms: overestimation, which overrates one’s own performance; overplacement, which rates oneself above others; and overprecision, which is excessive confidence that one knows the truth. Forecasts and budgets show overprecision directly, because a single number or a narrow range states a confidence level.

Every forecast carries an implied confidence. A sales target with a narrow band says the team is very sure. The implied confidence can be written as a percentage and checked later, which is what turns a vague trait into a measurement.

How do you build a calibration log for executive forecasts?

A calibration log has 4 columns: the forecast, the stated confidence as a percentage, the due date and the outcome. Executives write the confidence at the time of the forecast, using values such as 60, 70, 80 or 90 percent. Entries take one minute each, and a log of 40 forecasts builds up in a quarter for a team that forecasts weekly.

Score each entry when its due date arrives, with a simple right or wrong. A forecast counts as right if the claimed outcome occurred or if the actual value lands inside the stated range. Keep the original entry unchanged. The unchanged record is what makes the later comparison honest.

What does a calibration table reveal about executive confidence?

Group the 40 illustrative forecasts by stated confidence, 10 per group. At 60 percent, 5 prove right, or 50 percent. At 70 percent, 6 are right, or 60 percent. At 80 percent, 6 are right, or 60 percent. At 90 percent, 7 are right, or 70 percent. Each group lands below its stated level.

Across the 40 forecasts the average stated confidence is 75 percent and the hit rate is 24 of 40, or 60 percent. The calibration gap is 15 points. A gap near zero means the confidence of the team can be read at face value, and a gap of 15 points gives a correction to apply to future claims.

How does a calibration score change a go decision?

Suppose a team states 85 percent confidence that a launch hits its revenue target. The investment is 100 and the payoff is 160 if the target is hit and 0 otherwise (US$ thousands, illustrative). At 85 percent the expected value is 0.85 × 160 minus 100, or 36. The number looks comfortable and invites an immediate yes.

The table says an 80 percent claim has proven right 60 percent of the time and a 90 percent claim 70 percent of the time, so an 85 percent claim maps to about 65 percent. The corrected expected value is 0.65 × 160 minus 100, or 4. The decision stays positive and the margin shrinks from 36 to 4, which sets a more modest budget and a closer watch.

How do wider ranges correct overprecision in an executive forecast?

Overprecision shows when ranges are too narrow. In the illustrative log, a team gave 80 percent ranges for 20 monthly sales figures, and 12 of the 20 actual values fell inside, a capture of 60 percent. The ranges promised 80 percent and delivered 60 percent, so the team was behaving as if it knew more than it did.

Widen each range by 50 percent around its midpoint and recount. In the illustration the capture rises from 12 of 20 to 16 of 20, or 80 percent, which matches the promise. The wider range is a more honest statement of uncertainty, and downstream plans for stock, staffing and cash can then use the wider band.

How often should executives refresh their calibration score?

Refresh the table every quarter and wait for at least 10 forecasts per confidence level before trusting a bucket. With fewer entries a single surprise swings the result. A group of 40 forecasts is a starting point, and a year of weekly forecasts adds 52 entries and tightens the picture.

Score the team and each member separately, and share the results as a learning tool. People who see their own gap can adjust the next claim directly, and a team table shows where a shared habit needs attention. Pair the table with a rule: any claim of 90 percent needs a written reason and a second opinion.

How does the EQOS Asia decision engine add a score to executive judgment?

A calibration table corrects stated confidence after the fact. The EQOS Asia engine adds a measurement made before the debate: the Audit pillar scores a system on 9 structural dimensions and returns φ(S) between 0 and 1, with the STOP zone below 0.618 and the OK zone above (EQOS Asia method page, as of October 2026).

The score comes from a deterministic core, so the same inputs return the same figure and a team can compare its confidence with a computed reading. A reader can try a first score in the simulator, and the 30-minute pre-diagnostic, scheduled in Singapore time (SGT), checks a real forecast against the 9 dimensions.

Frequently asked questions

About the author

Stéphane Jambu

President, EQOS Forecast Asia

Stéphane Jambu leads EQOS Forecast Asia and works as a consultant in e-commerce, search visibility, and AI. LinkedIn profile