What it is
Overconfidence manifests in three related forms: overestimation of one's own performance, overprecision in the confidence intervals placed around estimates, and the tendency to rank oneself above average on relevant skills. All three are well-documented in managerial populations. Calibration studies consistently show that when people say they are 90 per cent confident in a range estimate, the true answer falls outside that range roughly 40 per cent of the time. Expertise reduces overconfidence in narrow technical domains but often increases it in judgements about broader outcomes.
Where it shows up
In build decisions, overconfidence is pervasive in the translation from technical specification to delivery estimate. Teams with genuine expertise in their domain systematically underestimate the complexity of integration, the volume of edge cases, and the capacity drag from operational and support work running in parallel. The result is not incompetence - it is overconfident calibration applied by capable people who have a limited reference class for the full delivery cycle.
What Rubicon Probity does
When Rubicon Probity processes a decision record at the Decide stage where confidence intervals are absent from key estimates - timeline, cost, adoption curve - it raises a CAUTION flag and requests that the team produce explicit confidence ranges rather than point estimates. The record must show a realistic range and document the assumptions that, if wrong, would cause the estimate to exceed the upper bound.
Detection questions
- Are the key estimates in this decision presented as point estimates or as ranges with documented confidence levels?
- What are the two or three assumptions that, if wrong, would cause the largest deviation from the central estimate - and how likely is each to hold?
- Has someone with no stake in this decision reviewed the estimates and confirmed they are calibrated against comparable external evidence?