The Nine-Box Grid for Small Companies: Running a Fair Talent Review
By Career Ladder Builder

Why a talent review at 80 people goes wrong in the first ten minutes
It is the second week of the planning cycle, and the leadership team of an 80-person company has booked ninety minutes to "do the talent review." The CEO shares a screen with a nine box grid drawn in a slide deck. Names have been typed into boxes in advance, mostly by the people who manage them. Within ten minutes the pattern is visible. The engineer who shipped the last release is a "high performer, high potential" because the VP of Engineering likes working with her. The operations lead who quietly rebuilt the onboarding process sits in the middle because nobody in the room has seen his work. A salesperson who missed quota twice is parked in the bottom-left corner after one sentence of discussion, and nobody asks what evidence supports it.
Nobody in this meeting is acting in bad faith. They are doing what small leadership teams do when they have a framework for sorting people and no method for filling it in. The grid gets populated from memory, recency, and personal rapport, and the output then feeds succession plans, stretch assignments, and promotion shortlists.
This article shows you how to set up the grid, how to rate performance and potential from separate evidence, how to run the meeting, and how to record what the team decided, so the result can be explained to the people on it.
How the nine box grid works, and what each box should trigger
The grid has two axes. The horizontal axis is performance: how well a person is doing the job they hold now. The vertical axis is potential: how likely a person is to succeed at a larger or different role. Each axis has three bands (low, moderate, high), which produces nine boxes.
The grid is only useful if each box leads to a different action. A workable set of default actions looks like this:
- High performance, high potential: Stretch assignment, sponsorship, and a named development plan. Confirm they know the company sees them this way.
- High performance, moderate potential: Retain and recognize. These people often anchor the team and may be right where they want to be.
- High performance, low potential for the next level: Deepen expertise in the current role. This is a statement about the next level, not a judgment of the person.
- Moderate performance, high potential: Find out what is blocking results. Often it is role fit, onboarding, or a manager problem.
- Moderate on both: Solid contributors. Keep them developing at a steady pace.
- Low performance, any potential: Start a performance conversation with specific, documented expectations before anything else.
The point of the exercise is the action column, not the placement. A box with no action attached is decoration.
Rate performance and potential separately, and from different evidence
The most common error in a nine box grid is letting one axis leak into the other. A strong performer is assumed to have high potential because the results are visible. A quiet performer is assumed to have low potential because the results are hard to see. Keep the axes apart by giving each its own evidence.
Performance should come from evidence about the current role: results against agreed goals, behavior against the competencies for the person's current level, and specific examples. If you already score employees on a 1–5 scale against level competencies with evidence notes, the current-level average is your performance input. Our guide to writing evidence notes covers what a usable note looks like.
Potential is harder, and it should never be a gut feeling. Define it as observable behavior that predicts success one level up. Four questions work for most small companies:
- Does the person learn quickly when the work changes?
- Do they take on problems outside their assigned scope, and finish them?
- Do others seek out their judgment, including people outside their team?
- Have they shown the specific competencies of the next level, even occasionally?
Rate each question on the same 1–5 scale with a written example. Questions three and four can be checked against the next level's expectations, which is the same comparison described in how to evaluate employee career readiness. Where the person is short of the next level, the gap is a development target, and a skill gap report is the natural way to list it.
A worked example, using round numbers for illustration: a senior analyst averages 4.2 against her current-level competencies. On the four potential questions she scores 4, 3, 2, and 3, an average of 3.0, with a note on each. She lands in the high-performance, moderate-potential box. That placement is defensible because anyone can ask which behavior produced the 2 on cross-team judgment and get an answer. It also points to a specific development action instead of a vague label.
Why insist on this discipline? Because individual raters are less reliable than they feel. In a widely cited study of the idiosyncratic rater effect, the individual rater's perceptions accounted for more than twice as much of the variance in performance ratings as actual performance did (Scullen, Mount & Goff, Journal of Applied Psychology, 2000). A grid filled in by one manager's impression inherits that problem, and a grid filled in by a room of managers can amplify it.
Prepare the evidence before anyone enters the room
A talent review meeting should not be where ratings are first formed. Two weeks before, ask each manager to submit for every direct report: a performance rating with evidence notes, the four potential ratings with an example for each, and a proposed box. Managers who cannot supply examples for a rating should say so, and that gap is useful information in itself.
Send a pre-read to the other leaders with the proposed placements. People who have worked with an employee outside the manager's team can add evidence in advance, which is the single best defense against a manager's blind spots. This also keeps the meeting from turning into a first reading of everyone's opinions.
Limit the scope. In a company of 30–200 people, review the full population only if the leadership team is small enough to know most people. Otherwise, start with managers, critical roles, and anyone being considered for promotion or a stretch role. A review that tries to cover everyone in ninety minutes ends up covering no one with care.
Run the meeting so the evidence leads
Use a fixed agenda and keep to it:
- Ground rules (5 minutes). State that the grid describes the current evidence, not a permanent label, and that it will be revisited.
- Calibrate on one or two cases first. Pick two people whose placements are uncontroversial and talk through the evidence. This sets the standard for what counts as support.
- Go person by person, evidence first. The manager presents the ratings and the examples before announcing the box. Then others add or challenge. Ask "what did you see?" before "what do you think?"
- Challenge placements that rest on adjectives. "Great attitude" and "not strategic" are not evidence. Ask for the behavior behind the word.
- Capture dissent. If two leaders disagree, record both views and the evidence behind each, rather than forcing a vote that hides the disagreement.
- Close with actions. For each person discussed, name the action, an owner, and a date.
The facilitator should not be the person who manages the most people in the room. In a small company that is often the head of People or an outside colleague. Their job is to ask for evidence, keep the discussion to the agreed criteria, and notice when the same two voices are carrying the room.
Watch for four recurring biases. Recency: the last two months outweigh the last year. Proximity: people who sit near leadership, or work in the same office, rate higher. Similarity: leaders rate people who resemble them more generously. Halo: strength in one visible area spills over into everything else. Naming these before the meeting makes them easier to call out during it. Rating consistency across managers is a separate problem that a calibration session addresses directly.
Check the outcome for fairness before it becomes a decision
Once the nine box grid is complete, look at it as a whole before it feeds any decision. Who is in the high-potential row? Is it concentrated in one team, one manager's reports, one demographic group, or one office? A skewed grid is not proof of bias, but it is a prompt to go back to the evidence for the people who sit just below the line.
This matters because the grid tends to feed promotion and development decisions. The EEOC's Uniform Guidelines on Employee Selection Procedures (29 CFR Part 1607) cover "hiring, promotion, or other employment decision," not hiring alone, and describe the four-fifths rule as a general indicator of adverse impact. A selection procedure with adverse impact is considered discriminatory unless the employer shows it is job-related and consistent with business necessity. This is general information, not legal advice. Confirm how these rules apply to your situation with employment counsel or the EEOC. For the practical side of the problem, see our piece on how to reduce promotion bias.
A placement you cannot explain with evidence is an opinion, and an opinion should not decide who gets developed.
Also tell people what the process is. You do not have to share a person's box, and many companies choose not to, but employees should know that talent reviews happen, what they are based on, and how to ask what would move them toward a larger role.
Log the actions and revisit them
The grid itself decays quickly. What lasts is the record of what you decided and why. For each person, keep the ratings, the evidence summary, the box, any dissenting views, the agreed action, the owner, and the review date. Revisit the log at the next review: did the actions happen, and did the evidence change?
Without a log, the next review starts from nothing, and the same impressions get re-entered. With one, you can see whether stretch assignments were delivered and whether a person's placement moved for reasons you can name.
If you want a ready-made structure for this, the Talent Review & Nine-Box Facilitation Pack provides a leadership talent review agenda, potential-calibration prompts, dissent capture, and a talent-action log around the 9-box grid. CareerEvaluations covers the layer underneath: scoring employees against level competencies on a 1–5 scale with evidence notes, and generating the gap reports that feed your potential ratings.
For more practical guidance on career frameworks, evaluations, and talent reviews, subscribe to the newsletter.
Enjoying this? Get more HR development guides in your inbox.


