How to Use Base Rates in Decisions
Start a forecast with outcomes from a relevant class of similar cases, then adjust for specific differences supported by evidence.
Example: on-time project delivery
Similar integrations finished on time 40 percent of the time. A dedicated team and proven connector justify an upward adjustment, but not a jump to 95 percent without outcome evidence.
Base rates anchor the story
A compelling plan explains why this case is special, but most cases are less special than they appear. The reference class supplies a starting frequency before unique details move the estimate.
Build a usable reference class
- Define the outcome and time window.
- Choose cases similar on causal features.
- Calculate the observed rate.
- Adjust for documented differences.
The wrong class misleads
A broad average may hide scale, regulation, or experience differences. Show why the class is relevant and use a range when selection is uncertain.
The decision rule
What to record before acting
Anchor on comparable outcomes before allowing case-specific evidence to move the forecast.
Apply the rule in context with practice probability judgment. Keep the evidence and assumptions with the decision so an outcome review can distinguish reasoning from luck.
Continue within evidence and uncertainty
Probability Range vs Point Estimate
Use a probability range when evidence supports a region but not a precise number, and show which assumption controls its width.
Probability Range vs Point EstimateHow Confidence Intervals Inform a Decision
Use a confidence interval to understand sampling uncertainty, not as the probability that the true value sits inside one computed interval.
How Confidence Intervals Inform a Decision