← HR Metrics·Compensation & Benefits
Pay Coefficient of Variation
Relative dispersion of a pay distribution — std dev as a percent of the mean — comparable across jobs of very different magnitude. Flags anomalously high internal pay variability / compares survey-source spread.
How it’s computed
(stddev / mean) * 100
What the evidence shows
Evidence (effect sizes, priors, validity) is syncing from Principia.
What this metric can show you
Pay Coefficient of Variation can tell roughly 22 pre-built stories — each a designed scene the data either confirms or it doesn’t. Bring your numbers and the Story Finder runs every one of these shapes against them.
specific to compensation & benefits
A real gradient — now ask if it's pointed at value
compensation · T1
Below the market, across the board
compensation · T1
One group sits apart on a decision that should be neutral
fairness-equity · T1
Pay is drifting from plan
compensation · T1
universal shapes — any single metric can take these
A few large values are doing the talking
any focus · T1
A one-time event, not a trend
any focus · T1
It doesn't track — the premise is false
any focus · T1
It's concentrated — one group stands apart
any focus · T1
Scenes are pre-built; your data is the toggle. Browse the full deck or watch one play end-to-end in The Quiet Exodus.
Run it on your data
This metric is computed in the People Analytics Toolbox on your own numbers. See pricing — posted, no quotes.
sources: toolbox:metrics-catalog
What the literature says
The measurement literature behind this signal — sourced, so you can defend it.
“Table 5.14 is our completed table with all four measures of variation we have discussed. Interpretations and Applications of P90/P10 1. The Values Themselves . The smaller the P90/P10, the smaller is the spread of the middle 80% of the data. The larger the P90/P10, the larger is…”
— Statistics for Compensationmatch 63%
“Similar calculations were done for the other ratings in Company A and for all the ratings in Company B. All values are rounded. Suppose it has been decided to use the 5-level system of Company A. Now we must convert the 1–7 scale of Company B to a 1–5 scale. The scheme we will…”
— Statistics for Compensationmatch 63%
“However, if you want to gain an understanding of data, there is no shortcut. You must measure variability and track it to build up your own trends and rules of thumb so that you not only gain an understanding of the data and underlying situations but also are able to identify…”
— Statistics for Compensationmatch 62%
Resources: Statistics for Compensation