Nine-Box Placement

Performance crossed with potential, one placement per worker per cycle. The grid is a snapshot; the movement between snapshots is where the decisions live.

RuleOne placement per worker per cycle, both axes from the same completed review, banded through the scale's configured sort.Neverread a small population's grid as signal. Nine cells over a small team is noise; suppress cells below a floor and roll up.

◆ The traps— per-cycle snapshots, banding both axes, calibration finality, and small-cell noise.

Trap	What goes wrong
Placements are cycle results	The same rule as ratings: one row per worker per cycle, never overwritten. Movement between cycles, who rose, who slid, is the actionable list, and it only exists if the history does.
Two axes, two configured scales	Performance and potential each ride their template's scale. Both band to low, medium, and high through dim_rating's numeric sort; banding by label text breaks the first time a template changes.
Completed placements only	Draft placements move in calibration sessions. The grid reads completed reviews; the in-progress view is monitoring, labeled.
Small cells are noise	Nine cells across a small team produces counts of one and two that read as insight and are chance. Set a floor, suppress cells beneath it, and report small organizations one level up.
The corners are the meeting	High-performance high-potential is the succession list; high-potential low-performance is a development conversation; low-low with tenure is the hardest one. The middle cell is most of the company and needs no meeting.

◆ The owned model, and the SQL— the grid from the review fact, and the proof it reconciles with the rating distribution.

No new fact. fct_review carries both ratings per cycle; the grid is one banded GROUP BY:

-- 2026 mid-year cycle: the grid from completed reviews
SELECT pot.band  AS potential_band,
       perf.band AS performance_band,
       COUNT(*)  AS workers
FROM fct_review r
JOIN dim_rating perf ON r.rating_key           = perf.rating_key
JOIN dim_rating pot  ON r.potential_rating_key = pot.rating_key
WHERE r.cycle_id = 'MY2026' AND r.review_status = 'Complete'
GROUP BY 1, 2

Sample output as the grid, the same 132 completed reviews as the Review Cycle report:

potential \ performance	Low (ratings 1–2)	Medium (3)	High (4–5)
High	2	15	12
Medium	6	40	26
Low	6	16	9

The columns sum to 14, 71, and 47: exactly the rating distribution from the Review Cycle report, banded. If the grid's performance axis does not reconcile with the rating report, one of them is reading drafts, and that is the audit. The 12 in the top-right corner are the succession bench; crossing the bottom row against the Turnover report's regretted exits, cycle over cycle, tests whether placements predict anything. Sample values are illustrative, never client data.

Want a nine-box with history that holds up?

Per-cycle placements, both axes banded through configured scales, and the movement list between cycles, reconciled against Workday, and you own every line.

Talk to us

Terms on this page

nine-box: Performance crossed with potential, three bands each. One placement per worker per cycle.
banding: Mapping a configured scale to low, medium, high through its numeric sort.
movement: Cell transitions between cycles. The actionable list the snapshots exist for.
cell floor: The minimum count below which a cell is suppressed as noise.
succession bench: The high-performance high-potential corner.
the reconciliation: Grid columns must equal the rating distribution, banded. If not, one side is reading drafts.