The original error: confusing performance with potential
Performance is what the person delivered in the current role, with evidence. Potential is the probability of taking on a role of greater complexity within a defined horizon. When both axes are filled with the same impression from the direct manager, the matrix becomes a perfect diagonal and loses any decision value.
Calibration is the product, not the matrix
The value sits in the session where several leaders defend, with evidence, why they placed each person in each box. That is where recency, affinity and generous-department biases get corrected.
- Written definitions for each axis, with observable behavioral anchors
- Calibration sessions by level, not by department, to force comparability
- Suggested distribution, never forced: rigid quotas destroy credibility
- A record of the evidence that justified each move between boxes
No consequences, no process
Every box must be tied to a concrete, budgeted action: succession plan, critical project assignment, compensation adjustment, development plan or exit conversation. If the box changes nothing in the next ninety days, the team learns the exercise is decorative and next year it will be filled on autopilot.
Where AI helps
People analytics models can contrast the leader's rating against objective signals — results, mobility, surveys, internal collaboration networks — and flag inconsistencies before the calibration session. The decision stays human; what changes is the quality of the evidence on the table.
Key takeaways
- The 9-box is worth the calibration session, not the completed matrix.
- Performance and potential require different definitions and evidence.
- Every box needs a budgeted action within 90 days.
NS frameworks and reference sources
- McKinsey — Talent management and organizational performance
- Korn Ferry — Assessment of potential frameworks
- Deloitte — Human Capital Trends
This article develops proprietary NS Business Strategy frameworks, drawing on our project base and on public industry literature and studies cited above. Figures are reference ranges; each project is measured against the client's actual data.