How Decision Confidence Works

Two different questions

"Which option is better" and "how sure should I be about this specific comparison" are not the same question. An 82-vs-78 result built entirely on contractual numbers (a stated salary, a fixed APR) deserves more trust than an identical 82-vs-78 gap built on optimistic guesses about an unpaid bonus or a rough resale-value estimate.

What lowers confidence

Confidence drops when inputs are incomplete, when a dimension is explicitly marked as an estimate or a subjective rating rather than a hard fact, when the result is fragile (a small change in one input would flip the winner), and when a scoring model is new enough that it hasn't accumulated much real-world track record yet.

Presented as a band, not a fake precise percentage

Confidence is shown as High, Medium, or Low rather than a bare number like "82%" — a raw percentage implies a level of statistical precision the underlying inputs don't actually have. The exact numeric index is still calculated and available if you want to see it, but it's not the headline, because presenting it that way would overstate how scientific the underlying estimate really is.

It never becomes a fabricated statistic

Confidence is never based on invented probabilities. A dimension marked "subjective" (like your own rating of a job's career-growth potential) carries a documented, fixed uncertainty discount — it isn't calibrated against real-world outcome data, because that data doesn't exist yet for any of these tools. That's disclosed here directly rather than implied away.

See a real Decision Confidence result using your own numbers.