The Calibration lens digs into how consistent your ratings actually are — across managers, against the firm average, and between self and manager perspectives.
What this page answers
How is a given rating dimension distributed? Are managers calibrating consistently? Where are people self-rating very differently from how their managers see them?
Pick a rating dimension
Calibration only makes sense for one rating dimension at a time. Use the sidebar dropdown to pick which dimension to analyse — it lists every rating dimension in the review that has at least five responses. Anything below five responses gets excluded because the numbers would be too noisy.
What’s on the page
Rating distribution
A bar chart showing the count of responses for each option on the rating scale. The chart title shows the dimension label and the total response count.
The shape of this bar chart tells you the calibration story at a glance:
Bell-shaped, centred near the middle: healthy calibration
Right-skewed (everyone rated highly): rating inflation
Left-skewed (everyone rated low): either a genuine problem cohort or harsh calibration
Bimodal: reviewers are split, often because of unclear scale anchors
Rating by manager
A table with one row per manager:
Manager: name
Reviews: team size / response count
Avg Rating: the manager’s team average on this dimension, to two decimals
vs Org Avg: a trend arrow showing whether the manager is above, below, or in line with the firm average (±0.3 threshold for “stable”)
Min: lowest rating they gave
Max: highest rating they gave
This is where rating outlier managers surface — anyone consistently above or below the org average is a candidate for a calibration conversation.
Self vs manager gap
A dumbbell chart showing every “paired” dimension — i.e., every competency that exists as both a self-rating and a manager-rating. For each pair, you see the Self and Manager averages side by side, sorted by the magnitude of the gap.
Pairs at the top of the chart are the dimensions with the biggest self-vs-manager disagreement. These are conversation-worthy: the firm is calibrated differently on these competencies than employees see themselves.
Where the data comes from
Calibration is the least AI-driven page in the Reviews tab. The data is the literal review responses:
The bar chart aggregates response counts
The manager table averages numeric ratings per manager
The paired-dimensions chart cross-references self-rating and manager-rating versions of the same competency
The only Compass involvement is the small “Analysis by Compass AI” attribution line — and even that’s really about the page existing as a structured calibration analysis, not about AI scoring.
How to read it
Start with the distribution. Is it healthy? If not, that tells you whether the issue is firm-wide or manager-specific.
Use the manager table to find outliers. Sort by Avg Rating to see who’s furthest from the average; look at Min and Max to see range.
Self vs manager gap is the most conversation-driving chart. If “Strategic Thinking” shows a wide gap with employees rating themselves much higher than managers, you’ve got a calibration conversation to have at firm level — not just with individuals.
What “paired dimensions” means
A paired dimension is a competency that exists twice in the review — once asking the employee to rate themselves, once asking the manager to rate the same competency. The page detects these pairs automatically from the dimension naming pattern (typically “X — Self Rating” and “X — Manager Rating”). If your review doesn’t include paired versions, this chart will be empty.
Tips
Pick the dimension that matters most for compensation. Calibration matters most for the dimensions that drive pay decisions — start there.
Compare cycle-over-cycle. A distribution that’s tightening across cycles is a sign calibration is improving; a distribution that’s spreading is the opposite.
Talk to managers, not the system. Calibration outliers are people-problems, not data-problems — surface the chart, but the fix is conversation and training.