Same scale, different managers, different bars

Sales rates 42.7 percent of its people at four or above. Technology Delivery and Support sit near 33. Either one group outperforms, or one bar sits lower, and an average cannot tell you which. The distributions and the outcomes join can.

RuleCompare rating distributions across manager groups on the same scale and cycle. A higher average is a question, not an accusation; the distribution is the evidence.

Neverread a group’s higher ratings as better performance without checking outcomes. Leniency and excellence look identical in an average.

◆ The same scale, three groupsaverage and share of top ratings, by department.

Manager group	Rated	Average rating	4s and 5s %
Sales	124	3.37	42.7
Technology Delivery	96	3.24	33.3
Support	76	3.24	32.9
All completed reviews	296	3.29	37.2

Sales rates 42.7 percent of its people at four or above; the other two groups sit near 33. Either Sales performs better or Sales rates easier, and the outcomes page is how you tell: if its high ratings do not predict better retention and promotion, the bar is lower, not the talent higher. Sample values are illustrative, never client data.

◆ The full distributionsthe shape behind each average.

Rating	Sales	Technology Delivery	Support	All
5, exceptional	8	6	4	18
4, exceeds	45	26	21	92
3, meets	58	52	41	151
2, developing	11	9	9	29
1, below	2	3	1	6
Rated	124	96	76	296

Every column sums to its group and every row sums to the published distribution, so this cut and the company view cannot disagree. The calibration page shows the room already pulled 31 ratings down; this table shows where the remaining spread lives.

◆ The owned answerone query, grouped by anything.

Distribution by manager group is the distribution query with one more column. The same cut works by tenure, by grade, or by any dimension the star carries, which is what makes the leniency conversation cheap to have and hard to dodge.

-- distribution and average by manager group
SELECT o.department,
  COUNT(*)                          AS rated,
  ROUND(AVG(r.final_rating), 2)     AS avg_rating,
  ROUND(SUM(CASE WHEN r.final_rating >= 4 THEN 1 ELSE 0 END)
        * 100.0 / COUNT(*), 1)      AS top_share_pct
FROM fct_rating r
JOIN dim_org o ON o.org_key = r.org_key
WHERE r.cycle_id = '2026-ANNUAL' AND r.status = 'Completed'
GROUP BY 1

Use case

Problem

One team’s ratings run high for years, comp follows the ratings, and nobody can say whether it is performance or a lower bar.

What we build

Rating distributions cut by manager group on the shared dimensions, reconciled to the company distribution, with outcomes joined for the verdict.

What you get

A leniency conversation with evidence: which groups rate high, whether their high ratings predict anything, and a merit spend that follows performance instead of generosity.

Does one team’s bar sit lower than the others?

We build the distribution cuts that make leniency visible and checkable.

Talk to us

Terms on this page

leniency: A group rating systematically higher for the same performance.
manager group: Workers rolled up by the org that rates them.
top share: The portion rated four or above; the compression metric.
same-scale rule: Cuts compare only within one scale and cycle.
outcomes check: Joining ratings to retention and promotion for the verdict.