9-Box Talent Grid

What this tool does

CM8-265 is a working 9-box talent grid. You place each person on two three-point scales — performance against the outcomes agreed for their role, and potential to grow into bigger or broader work — and the tool names the box, draws the grid, tracks flight risk and succession cover, and prints a report for the talent-review meeting. One row per person per review cycle, so the file becomes a history of how your view of each person has moved.

The tool insists on the discipline that makes a 9-box worth doing: top-box placements will not save without written evidence, a bottom-left placement will not save without an agreed action, and your best people at high flight risk will not save without a retention action.

Tiles, charts and summary tables all read the latest cycle — the cycle whose rows carry the most recent review date — so older cycles stay in the register as history without polluting this review's numbers.

What the 9-box is — and is not

The 9-box is a conversation structure, not a verdict. Its value is that it forces a room of managers to compare their judgements against the same two questions and defend them with evidence. The moment a placement becomes a label that follows a person around — "she's a 2,2" — the tool has failed. Boxes describe a person in this role, this cycle, on this evidence; people move between boxes constantly, and a grid where nobody ever moves is a grid nobody is honestly re-scoring.

One honest paragraph about confidentiality, because it matters more here than in almost any register: this file contains judgements about named, identifiable people — judgements that would damage trust and morale if they leaked, and that in many countries the people concerned have a legal right to see. That is precisely why this tool runs entirely in your browser: there is no account, no upload and no network request of any kind, so the placements never leave the computer you are using. The flip side is that you are the security: keep exports off shared drives, use initials if the file must circulate, and never leave the grid open on a screen in an open office.

Performance vs potential — observable anchors

The two axes are different questions and must be scored from different evidence.

Performance is backward-looking and comparatively easy: against the outcomes agreed for the role, over the whole cycle, did this person miss (1), deliver (2) or consistently outperform (3)? The traps are recency — scoring the last month, not the cycle — and difficulty blindness, where the person with the hardest brief scores lower than the person with the easiest one.

Potential is forward-looking and where reviews go wrong. It means the capacity to succeed in bigger or broader roles: limited — right level now (1); growth — one level up or broader scope (2); high — two levels or a leadership track (3). Anchor it in things you have observed: learns new domains fast, is sought out by peers for judgement, performs when the situation is ambiguous, lifts the people around them. Three classic rater errors to name out loud in every session:

• Potential is not ambition. Wanting the bigger job is not evidence of being able to do it — and not wanting it (yet) is not evidence of a ceiling. Quiet people are systematically under-scored.
• Potential is not likeability. "Great attitude" and "everyone loves working with him" are congeniality, not capacity. Demand an example of the person succeeding at something above their level.
• Potential is not performance repeated. The best salesperson is not automatically the next sales manager. High performance in this role tells you about this role.

Running a calibration session

Placements made by one manager alone measure that manager as much as their people — some rate hard, some rate kind. Calibration is how you correct for it: managers propose, peers challenge, evidence decides.

1. Before the session, every manager enters draft placements with the evidence written down. The tool's evidence field is the ticket to the meeting — no evidence, no discussion.
2. In the session, walk the grid box by box, not manager by manager. Put every proposed Star on the table together and ask: would each of these people be a Star on any team in this room? Do the same for the 1,1s.
3. Anyone may challenge any placement, but only with evidence — an observed example, an outcome, a comparison. "I just see her differently" is an opinion; "she led the recovery when the release failed and his equivalent situation went the other way" is calibration.
4. When the room agrees, mark the row Calibrated. When the development action is agreed and owned, mark it Actioned. The calibration-progress tile shows how far the cycle has actually got.

The boxes that matter most

Three placements deserve most of the meeting's time:

• Enigma (performance 1, potential 3) — high capability, low delivery. Almost always one of two things: a new hire still landing, or a good person misplaced, badly managed or disengaged. Either way the answer is the same: decide fast. Set a short, explicit reset with support and a clear bar, and re-place the person at a mid-cycle check. An Enigma left to drift for a year becomes a resignation or a performance case — both avoidable.
• Star at flight risk — your most expensive vacancy, currently still on the payroll. Replacing a genuine Star costs a year and a multiple of salary, and the market knows what they are worth even if your pay bands do not. The watchlist table exists for exactly this row: retention action, named owner, this week.
• Underperformer (1,1) — the box organisations archive instead of acting on. A 1,1 with no action is just a label that everyone else on the team can see being tolerated. Act with dignity: an honest conversation, a real improvement plan with support and a deadline, and if the bar is not met, a fair exit. The tool will not save a 1,1 without an action for this reason.

Development actions by box family

Different box families need different kinds of action — a development plan that says "keep it up" for everyone is a form, not a plan.

• Top talent (Star, High potential, High performer): stretch, exposure and retention. Bigger assignments before bigger titles, a mentor above their level, visibility to senior leaders — and for anyone at flight risk, a retention conversation that happens before the resignation, not after.
• The middle (Core player, Effective, Solid professional): mastery and respect. These people deliver most of the output; the worst thing you can do is treat the middle as a waiting room. Deepen craft, broaden modestly, and be honest that "valuable at this level" is a compliment.
• The left column (Enigma, Inconsistent, Underperformer): diagnosis before development. Is it role fit, management, skill, will or circumstance? Time-boxed support with an explicit bar, and a decision at the end of it either way.

The centre-box trap

When raters are unsure, unprepared or conflict-averse, everyone lands in Core player — it feels safe, it offends nobody, and it requires no evidence. A grid where most of the population sits in the centre box is usually a rating failure, not a workforce shape. The centre-box tile flags the share above your chosen threshold (half, by default). When it fires, the fix is not to force-rank people out of the middle — forced distributions create their own injustices — but to ask, person by person: what evidence puts this person here rather than one box in any direction? A genuine 2,2 with written evidence, like a steady engineer delivering exactly what the role asks, is a perfectly good placement. Ten of them with blank evidence fields are ten conversations that never happened.

Succession linkage

The grid and the succession plan are two views of the same risk. Every Star, High potential and High performer is someone whose departure would hurt — so each of those rows carries a simple checkbox: is a successor identified for their role? The succession-coverage tile turns it into one number. Two patterns to watch: a high performer with no successor is a single point of failure you already know about; and your High potentials are themselves the successor pool — the cross-training action that covers one person's succession risk is often another person's development plan. Wire them together deliberately.

Cadence

Twice a year is the rhythm that works for most organisations. Annually, the grid goes stale — people join, leave, and change more than one box's worth inside twelve months. Quarterly, the process eats more management time than it returns and placements barely move. Run a full review with calibration every six months, keep the cycle name in the register ("2026 H1", "2026 H2"), and use a light mid-cycle check only for the rows that demanded it: Enigmas on a reset, 1,1s on a plan, top talent at flight risk. Because every cycle keeps its rows, the register becomes the record of movement — and movement over two or three cycles is far more informative than any single placement.

FAQ

Should people see their own box? Share the substance, not necessarily the label. Everyone deserves an honest conversation about how they are doing and what is next; whether the phrase "9-box" or the box name helps that conversation depends on your culture. What is never defensible is a grid whose subjects experience its consequences without ever hearing its content.

How big a population fits one grid? Calibration works when the room knows the people — roughly 15 to 60 per session. Bigger organisations run one grid per function and a second-level review above.

Can someone be a 3 on potential twice running without moving? Once, yes. Twice, the question moves to you: if they are genuinely ready for more and nothing has been offered, the flight-risk field is about to answer for you.

Is a 1,1 placement grounds for dismissal? No — it is a judgement in a management tool, not a process. Employment law on managing poor performance differs by country; the placement tells you a formal, properly-documented process needs to start, run by whoever owns that process in your organisation.

Why only three points per axis? Because the conversation, not the scale, does the work. Five-point axes produce false precision and longer arguments about 3-versus-4; three points force the only distinctions that change what you do next.

Saving your work

Placements, settings and the report header are written to this browser's local storage as you type, and the toolbar shows the time of the last save. That storage belongs to one browser on one computer: another browser, a private window, a second machine or a clean-up tool that clears site data will not have it.

Treat Export .json as the real save — one file containing everything, which Import .json restores anywhere. Export CSV gives you the register for spreadsheet work. Reset asks twice, then erases everything this tool has stored. There is no undo. This file is confidential personal data about named people — store and share exports accordingly.

Accuracy & disclaimer

The arithmetic here is trivial and the tool does it faithfully; everything that matters sits underneath it. A placement is a manager's judgement made comparable by calibration — it is not a measurement, and the grid ranks the quality of your evidence as much as it ranks your people. Two organisations' grids are not comparable, and neither are two cycles scored to different standards.

Nothing in this tool is employment-law, HR-process or compensation advice. How performance may be managed, what records employees may access, and what consultation is required before decisions that affect people all differ by country — take advice that applies to you before acting on any placement, especially at the corners of the grid.