PlantOps BI

Rate Comparison

Enter the incident count and hours worked for two periods. You get each rate per 200,000 hours (the TRIR basis), an exact 95% range for each, and the chance of a split at least this uneven if nothing had changed.

Earlier period
Later period
Earlier rate
-
Later rate
-
p-value
-

Enter both periods.

How to Read It

The p-value is the probability, if the true rate were the same in both periods, of seeing a split of incidents at least this lopsided given the hours in each. Small values (under 0.05 is the usual line) say the change is unlikely to be chance alone. A large value does not prove nothing changed; with a handful of incidents there is often too little data to tell.

Don't judge from whether the two ranges overlap. Ranges can overlap while the difference is still significant, which is why the tool runs the test directly. The defaults are the demo data's Boise plant in 2024 and 2025, from the TRIR with small numbers analysis, which also has the DAX for these ranges.

Ranges are exact (Garwood) Poisson limits; the test conditions on the total count, so the later period's share of incidents is binomial with its share of hours. The site's tests check both against published values and hand calculations.