Engineering, production, governance, cost, communication · insight-1295
Replace probability words with numeric ranges in written risk statements
- Evidence type
- practitioner heuristic
- Difficulty
- easy
- Roles
- Data scientist, Data analyst, Applied AI
- Industries
- cross-industry, public-sector, finance
Summary: situation, action, reason, boundary
Situation
A risk memo calls a forecast regression 'likely' and a launch delay 'possible', and three readers take away different odds (hypothetical).
Action
Attach a number or range to every probability word, define a one-line legend (for example likely = 65 to 85 percent, a hypothetical convention), keep the legend identical across documents, and say when a number is a judgement.
Reason
In an early intelligence estimate, the same phrase was held by colleagues to mean odds from about 20 to 80 up to 80 to 20, which prompted a proposal for fixed numerical ranges for key words.
Boundary
Decision memos, risk registers, model review notes and forecast commentary.
Symptom or risk
Readers attach very different odds to the same word (Kent found one phrase read as anything from about 20 to 80 up to 80 to 20), so decisions rest on odds the author did not mean.
Exceptions
Informal discussion needs no numbers; if you cannot defend a number, say that.
Trade-offs
Numbers imply precision, so give ranges.
Pitfalls
A legend nobody reads, or numbers invented to fit the words.
Lessons and concepts
Claims and their sources
Kent recounts that colleagues who agreed an estimate using the phrase 'serious possibility' had in mind odds ranging from about 20 to 80 up to 80 to 20, and argues for consistent use of a few odds expressions with numerical ranges.