The Talent Grid is the most familiar PI view: a 3×3 grid with quadrants like Star, Core Performer, Flight Risk. Where it differs from a traditional manual grid is that every employee is placed automatically using Compass’s signals — no committee calibration needed up front.
What this page answers
Where does each employee sit on a two-axis map of (typically) performance and engagement? Which quadrants are full, which are thin? Who’s in the top-right and who’s in the bottom-left?
What’s on the page
The 9-box grid
A 3×3 visual matrix with cells colour-coded and labelled (e.g., “Star”, “Core Performer”, “Struggling & Unhappy”). Each cell shows:
The cell name
The count of employees in it
The percentage of the placed cohort
Cells are clickable — clicking one filters the employee table below to show only employees in that box.
Axis labels
The grid shows axis labels on all four sides with directional arrows. The X axis is “low → high” of whatever field you pick; same for Y.
Employee table
Below the grid is a sortable, paginated table of every placed employee:
Name (clickable for details)
Department
Grid position (which cell they’re in)
AI Performance level
If your X or Y axis is a rating dimension, the rating values appear as additional columns
Talent grid by department
A heatmap showing the count of employees from each department in each 9-box cell. Useful for spotting which departments are top-heavy, bottom-heavy, or evenly spread.
Configuring the axes
Use the sidebar to set the X and Y axes. The options come in two groups:
AI Extraction: Compass-produced signals: Performance Level and Sentiment
Submitted Ratings: any rating dimension from the review (e.g., “Leadership”, “Technical Skills”)
Defaults are usually Performance Level (X) and Sentiment Score (Y), giving the classic performance-engagement map. But you can equally plot Performance Level vs Leadership, or Manager Rating vs Self Rating.
How placement works
Each axis is bucketed into three levels (low / medium / high). The bucketing differs depending on the field:
Performance level: Low = Critical / Concerning; Medium = Solid; High = Strong / Exceptional
Sentiment score: Low = below −0.2; Medium = between; High = at or above 0.3
Rating dimensions: bucketed into three equal ranges across the scale (terciles)
Employees missing a value on either axis are skipped — you’ll see an “insufficient data” count in the page subtitle. This is usually because their review didn’t include enough data for the AI to score the relevant field.
Where the data comes from
The grid is populated entirely from review extraction — Compass reads each review and produces the underlying scores. Rating dimensions come straight from the review responses (no AI involved). Named 9-box labels are pre-set for the Performance Level × Sentiment combination (whichever way round you plot it); any other combination shows generic position labels (e.g., “High / Low”).
How to read it
Top-right (high/high) is your star quadrant. Strong performers who are also positive — your retention focus.
Bottom-right (high performance, low sentiment). Strong performers who aren’t happy — flight risks, deserve attention.
Top-left (low performance, high sentiment). Happy but not delivering — coaching opportunity.
Bottom-left. Struggling and unhappy — usually the action priority.
Middle cells are the bulk of any healthy cohort.
Tips
Don’t over-interpret single placements. A box position is one read on one cycle. Use the Employee Trajectory tab to see if someone’s moving across cycles.
Try swapping axes. Performance × Sentiment is the default, but Performance × Leadership Rating can be more revealing for a senior cohort.
The skipped count matters. A grid that’s only showing half the cohort isn’t representative. Investigate why — usually it’s extraction still in flight, or thin review content.
The department heatmap is where succession conversations start. A department with zero high-high employees has a real bench problem worth flagging.