Learn
A percentage is not automatically a probability
Two numbers often travel under the same percent sign. One is a probability, a claim about how often an event will happen. The other is confidence, a claim about how well the case for an answer has been tested. Reading one as the other is a common and expensive mistake.
What a probability promises
A probability is a forecast you can check over time. If a forecaster says 70 percent chance of rain on a hundred different days, it should rain on about seventy of them. When it does, the forecaster is called well calibrated, and the number has earned the right to be read as a frequency.
That check needs many similar events and an honest count of how they turned out. Weather offers both. Most business choices offer neither: a company opens its second plant once, and nobody gets to rerun the year.
What confidence describes
Confidence answers a different question: how much of the case has been tested, and how well did it hold. A recommendation resting on three sourced figures and a survived attack earns high confidence. The same recommendation resting on one unconfirmed estimate earns low confidence, even if it later turns out right.
So a low confidence answer is no coin toss. It is a fair warning that the record behind it is thin, and it tells you what to go and check.
How the confusion costs money
The usual error runs like this. A team sees 80 percent beside a plan, hears four chances in five of success, and commits money accordingly. In truth the 80 described how firmly the analysts agreed, and nobody had tested the sales figure the whole plan rested on.
The cure is to ask, every time a percentage appears, what it counts. If nobody can name the events it is a frequency of, treat it as a description of the analysis, whatever its symbol.
How a Verdict states its confidence
In Argumont, confidence describes the strength of the record, never the chance that the leading option is right. The Inspector marks each claim as supported, partly supported, unverified or contradicted, and the Judge lets those marks shape how firmly the Verdict speaks. When the record cannot carry a winner at all, the Verdict says No Strong Winner Yet.
Whatever the Verdict says, the Human Decision is still yours to make. Recording the Outcome later is how you learn whether your own judgment tends to run hot or cold.
Two readings of one number
A Verdict favors raising prices 8 percent, with moderate confidence. The record shows why: the cost figures come from signed quotes and are supported, but the estimate that no more than 5 percent of customers will leave is only the sales lead's guess, marked unverified. Reading moderate confidence as a 60 percent chance of success misses the point. The useful reading is narrower and more practical: test the churn estimate on one region for a month, and the confidence will either rise or the Verdict will change.
Questions
- Is a confidence level ever a probability?
- Only when it has been tested against many outcomes and found to match them. A single strategic choice rarely allows that, so the safer habit is to read confidence as a description of the evidence.
- Why not just give me a probability?
- Because a number invented for a one time choice would look precise and mean little. A plain statement of what the answer rests on, and what would change it, is more useful to someone about to spend money.
- How can I tell if my own estimates are well calibrated?
- Write your estimates down before events resolve, then count. If the things you call 90 percent certain happen only six times in ten, you are overconfident, and the tally will show it plainly.
Bring Argumont a Decision worth challenging.
Start with a paid 60 day pilot on your own real Decisions.