What is the 9-box grid?

The 9-box grid is a talent review tool. You place every person in a group onto a three-by-three grid — performance along one axis, potential along the other — and end up with a single picture of where the group's strength, risk and untapped capacity sit.

The grid is usually traced back to work McKinsey did with General Electric in the 1970s, where a similar nine-cell matrix was used to decide which business units deserved investment. The version HR teams use today applies the same logic to individuals. That attribution is widely repeated but thinly documented, so treat it as background rather than provenance — nothing about whether the grid is useful depends on where it came from.

Performance (horizontal) — what this person has actually delivered. Performance is backward-looking and evidence-based. It asks a narrow question: over the review period just ended, how did this person do against the expectations of the job they hold right now? Results, quality, reliability, how they worked with others. You should be able to point at something specific for every rating.

Potential (vertical) — how much larger or more complex a role this person could grow into. Potential is forward-looking and, unavoidably, a judgement. It asks a different question: if the organisation gave this person materially bigger or different work, could they do it, and do they want it? The useful signals are learning speed, how they behave when the problem is unfamiliar, whether their influence already reaches beyond their own remit, and — the one most often skipped — whether they actually want a bigger job.

Each axis has three levels: low, medium and high. That gives nine boxes.

The grid is not a scoring exercise. Its output is not really the nine boxes — it is a shared, comparable view of a population that several managers now agree on.

A typical cycle looks like this. Managers place their own people first, individually. Those placements are then brought into a calibration session where the managers, their common leader and an HR partner work through the grid together, and — crucially — managers have to justify their placements to peers who manage comparable people. That is where the value is created. A manager who rates everyone highly gets challenged. A manager who is unusually harsh gets challenged. Two people in different teams doing similar work get compared directly, often for the first time. Placements move during the conversation, and they should.

What comes out is a set of decisions: who needs a development action and what it is, who is a retention risk, which senior roles have no credible successor, and who needs a difficult conversation that has been postponed.

In succession planning specifically, the grid is an input and not the plan. What it gives you is a shortlist: the people with the capability and the appetite to take on materially larger work. What it cannot give you is readiness. High potential means someone could grow into a bigger role; it says nothing about whether they could do a named job next quarter. Keep readiness as a separate judgement — ready now, ready in one to two years, ready in three or more — recorded against the role rather than against the person. A succession plan that consists of the top-right box and nothing else is not a plan, it is a list of the people most likely to leave.

The most useful output is often the empty space. A high-potential row with nobody in it, in a team of forty, is a succession problem that was invisible before you drew the grid. A senior role with no credible successor anywhere on the page is the single most actionable thing a talent review produces, and it is the thing most likely to be skipped because it is nobody's individual development plan.

Read the step-by-step guide to running the session, including who should be in the room and what to do with the output — or go straight to the talent review FAQ if you have one specific question.

Anyone selling you the 9-box grid without this section is selling you something. The method has real, well-understood weaknesses.

Potential is the weak axis. There is no agreed definition of potential and no reliable way to measure it. In practice it often gets rated on proxies: confidence, visibility, articulacy, willingness to relocate or work long hours, or how much the person resembles the people doing the rating. Every one of those correlates with things that have nothing to do with capability.

Recency bias distorts performance. A review covering twelve months is usually rated on the last two. A great Q4 lifts someone a box; a bad project in the final month drops them one. Written evidence gathered through the year is the only real defence, and almost nobody does it.

The grid amplifies bias rather than filtering it. Structure creates an impression of objectivity, but a biased judgement placed in a nine-cell matrix is still a biased judgement — now with an official-looking position. Patterns worth actively checking before you accept a grid: how people returning from parental leave are rated, how part-time workers are rated, and whether "high potential" tracks demographics more closely than it tracks output.

Labels stick to people. The boxes are convenient shorthand, which is precisely the problem. Once someone has been described as a Core Player in a meeting, that description tends to follow them into the next cycle and into decisions they never see. Assume anything written on a grid will eventually be seen by, or repeated to, the person on it.

Forced distribution breaks it. Requiring a fixed percentage in each box turns a discussion into a rationing exercise, and managers respond by trading placements rather than assessing people. If a genuinely strong team produces a top-heavy grid, that is information, not an error to be corrected.

It is a snapshot, not a verdict. The grid describes one group at one moment, with the information available. People move between boxes, and a change of manager, role or brief moves them faster than anything else.

It does not measure several things people read into it. The grid says nothing about flight risk, about how critical a role is to the business, or about how soon someone would be ready for a promotion. "Star" does not mean "ready now". If you need those, track them separately — do not infer them from a box.

Used well, the 9-box grid is a structured excuse for a conversation that most organisations otherwise never have. Used badly, it is a way of writing down prejudices in a grid and calling the result data. The difference is entirely in how the session is run.

Every weakness above has a countermeasure, and none of them is complicated. They are just skipped.

Define both axes in writing before anyone is rated. Not during the meeting, and not in the manager's head. Write what low, medium and high mean on each axis, in your own organisation's language, and circulate it. If nine managers hold nine private definitions of high potential, the calibration session becomes a two-hour argument about vocabulary.

Rate the axes separately, and in that order. Do performance first and finish it, then start potential from scratch. The most common single error is letting a strong performance rating drag the potential rating up because it feels harsh not to. When that happens both axes say the same thing and the grid collapses into a ranked list you could have written without it.

Demand one piece of evidence per placement. "She led the migration and it landed two weeks early" is evidence. "He's very strong" is not. Requiring evidence in writing, before the meeting, catches a surprising proportion of gut-feel ratings before they reach the room.

Ask for evidence from the first half of the period. This is the only practical defence against recency bias that works in a live meeting. If every example a manager offers is from the last six weeks, you are not rating a year.

Compare people, not definitions. Arguing in the abstract about whether someone is medium or high goes nowhere. Naming two specific people and asking whether one is genuinely operating above the other resolves it in a fraction of the time.

Read the shape before you close. Step back and look at the distribution rather than the individuals. Is the top-right corner crowded — ratings inflation, most likely, rather than an exceptional team? Is the high-potential row empty? Does high potential track demographics, or volume, or which manager is most persuasive, more closely than it tracks output? Someone in the room has to be willing to ask that out loud, and it is usually the HR partner's job.

Attach an owner and a date to every placement. The grid is not the deliverable. A development action, a stretch assignment, a coaching conversation, a role change or a postponed difficult conversation — each with a named owner and a date — is the deliverable. Without that you have held a two-hour meeting and produced a picture.

Write it as though the person will read it. Because sooner or later someone will: a subject access request, a leaver's handover, a forwarded slide. Notes you would not stand behind in front of the person are notes that should not be on the grid.

The step-by-step guide turns each of these into a numbered step, with an agenda and timings for the session itself.