Protocol
Attach confidence to the evidence, not to your tone
Confidence should describe what is known, not how strongly the estimator speaks.
When it fits
- A date is delivered confidently even though key inputs are still unknown.
When to avoid it
- Ordinal confidence labels are editorial aids, not calibrated probabilities unless the team has defined and tested them.
Why it matters
State the evidence maturity behind the forecast: known scope, validated dependencies, available historical analogues and unresolved unknowns. Use a wider range or lower confidence when important inputs are still speculative, then narrow it when real uncertainty has been removed.
Steps
- A confidence statement names the evidence that supports it and what is still missing.
An example
A draft integration estimate stays wide until the external API and sample payload are tested.
Check your result
A confidence statement names the evidence that supports it and what is still missing.
Keep this limit in mind
- Ordinal confidence labels are editorial aids, not calibrated probabilities unless the team has defined and tested them.
Connected ideas
Useful withLet ranges narrow only when uncertainty is actually removed
Evidence and sources
Supports
IPA guidance notes that uncertainty and estimate range depend on the maturity and variability of input data, so estimates should evolve as design and evidence mature.
A later estimate can still be wrong; maturity narrows uncertainty only when real information has improved.
Cost Estimating Guidance · Estimate maturity and range