Purpose

A talent review exists to answer one question: does every person on the team have what they need to grow, stay, and do their best work?

It is not a ranking exercise. It is not a forced distribution. It is not a performance management tool. It is a structured conversation between managers - and their leadership - about each person's trajectory, their risks, and the actions needed to support them.

Done well, a talent review:

• Surfaces development needs that 1:1s consistently miss because neither party raises them
• Creates a shared, calibrated view of the team's capability across managers
• Identifies retention risks before they become resignations
• Generates a concrete action per person, owned by a named manager
• Builds alignment between managers who otherwise operate in silos

Done badly, it becomes a gossip session, a ranking exercise, or a document that sits in a folder until next year.

This playbook covers how to run it well.

When to Use This Playbook

Use this playbook when:

• You are running a structured talent review cycle (typically twice a year - mid-year and year-end)
• You have a team large enough that a skip-level cannot know every individual well
• You are preparing for a calibration session and need a consistent input format
• You have a new HR partner and want to establish a clear process
• You are running a talent review for the first time and want a structure to follow

This is not the right tool for:

• Ad hoc conversations about a specific individual (use a 1:1 or a direct conversation)
• Performance management of a struggling individual (use the performance management playbook)
• Promotion decisions (talent review informs these but does not make them)

Before You Start

Who attends

Keep the room small. Every additional person changes what managers are willing to say. If you have more than five people in the room, the honest conversation disappears.

What each manager prepares

Every manager brings a one-page summary per person. Not a slide deck. Not a 500-word essay. One page. It covers:

Current performance - how are they performing against their role expectations right now? Not potential. Not trajectory. Current delivery. One short paragraph.

Trajectory - are they growing, stable, or declining relative to their level? Growing means accelerating - doing more than you'd expect at this level. Stable means solid and consistent. Declining means something has changed and performance is dropping.

Retention risk - what is the risk that this person leaves in the next 12 months? Low, medium, or high. And what is the primary driver? Compensation, lack of growth, manager relationship, personal circumstances, external opportunity?

Development need - what is the one thing that would most unlock this person's growth? Be specific. "Better communication" is not a development need. "Presenting technical options clearly to non-technical stakeholders in architecture decisions" is a development need.

Next role - what is the likely next role for this person? Could be a promotion, a lateral move to broaden, or staying in role and deepening. If you cannot answer this, that is useful information.

What the facilitator prepares

The skip-level or HR partner prepares:

• A roster of every person being reviewed, grouped by manager
• The running order (how long per person - typically 5-8 minutes)
• A blank action log (person, agreed action, owner, due date)
• Ground rules for the session

Ground rules to share at the start

Read these aloud at the start of every session. Do not assume people know them.

1. Everything discussed stays in this room. Managers do not share what was said about specific individuals with those individuals unless it is part of an agreed follow-up.
2. We are here to help people, not to judge them. Assessments are working hypotheses, not verdicts.
3. If you disagree with an assessment, say so. Disagreement is useful. Silence is not.
4. Every person leaves this session with at least one action. "No action needed" is not an acceptable output.
5. We will challenge assessments that are not backed by specific examples.

The Process

Step 1 - Open the session (5 minutes)

Start by reminding the group of the purpose. Say something like:

"We are here to build a shared picture of where everyone is and what they need. This is not a ranking exercise. I am going to need you to back your assessments with examples - if I ask you what you mean, that is not a challenge to your judgement, it is how we make sure we are all talking about the same thing."

Confirm the running order. Confirm the time per person. Start.

Step 2 - Manager presents each person (3-4 minutes per person)

The presenting manager walks through their one-page summary:

• Current performance (one sentence)
• Trajectory (growing / stable / declining - and why)
• Retention risk (low / medium / high - and primary driver)
• Development need (specific and actionable)
• Next role (and rough timeframe)

They do not read the page aloud. They speak to it. If they are reading aloud, the summary is doing the wrong job.

Step 3 - Open for questions and challenge (3-4 minutes per person)

The facilitator opens the floor. The questions to ask:

• "What would change your view of their trajectory?"
• "When you say retention risk is high - what specifically are you seeing?"
• "What have you tried for that development need, and what happened?"
• "Does anyone in the room have a different view based on working with this person?"
• "What does the person themselves think about their next step?"

If the manager says "I think they're doing great" and cannot give a specific example of what great looks like, probe it. Not aggressively. But specifically.

Step 4 - Agree the action (1-2 minutes per person)

Every person gets an action. The format is:

• What: one specific action
• Who: named manager (usually the direct manager, sometimes the skip-level)
• When: a real date, not "Q3" or "soon"

Examples of good actions:

• "Arrange a conversation between them and [senior engineer] to explore the platform team role - by end of month"
• "Have a direct conversation about the compensation gap and what the path looks like - before the calibration session"
• "Discuss the feedback about stakeholder communication in next 1:1, share a specific example, agree a development goal - this week"
• "No growth path visible in current role - explore options for broadening scope - within 6 weeks"

Examples of bad actions:

• "Keep doing what we're doing" - this is not an action
• "Monitor" - monitor what, by when, and then what?
• "Have a development conversation" - when? about what?

Step 5 - Use the 9-box to structure thinking (optional but useful)

The 9-box plots people on two axes: current performance (low / meeting / exceeding) and potential / trajectory (low / medium / high). It is a tool for structuring conversation, not a verdict.

Use it as a starting point, not an output. The value is in the conversation about where someone sits and why - not in the label itself.

Limitations of the 9-box to name explicitly:

• "Potential" is notoriously biased. People from underrepresented groups are systematically rated lower on potential and higher on performance. Name this explicitly.
• The 9-box is a snapshot. It is not a character assessment. Someone in the bottom-left today can move to the top-right in two years.
• Managers often rate people they like higher on potential. Ask for evidence, not impressions.
• Do not use 9-box language with individuals. Telling someone they are a "low potential" is damaging and untrue. The output for individuals is actions and conversations, not labels.

Step 6 - Handle disagreement between managers

When two managers see the same person differently, do not paper over it. It is useful information.

Ask:

• "What specifically are you each seeing that leads to different views?"
• "Are you seeing them in different contexts - and could both be true?"
• "What would you need to see to update your view?"

If the disagreement cannot be resolved in the session, agree a method for resolving it: a conversation between the two managers and the person, a specific observation to make over the next month, or a follow-up in four weeks.

Do not average the disagreement into a middle rating. That satisfies no one and tells you nothing.

Step 7 - Close the session

At the end, the facilitator reads back all agreed actions. Confirm owner and date for each one.

Then ask: "Is there anyone we did not discuss who should have been on the list?" This catches people who were overlooked because their manager did not flag them.

Close with a reminder of confidentiality expectations. What was discussed stays in the room. The outputs (actions) will be actioned. The assessments (labels, ratings) will not be shared verbatim with individuals.

What Good Looks Like

A good talent review session:

• Runs through every person in the allotted time without feeling rushed
• Has at least three meaningful challenges to an initial assessment - where someone pushes back on a label and the group updates its view
• Produces a named, dated action for every single person
• Surfaces at least one retention risk that was not previously known to the skip-level
• Ends with managers feeling like they have a shared picture rather than five separate pictures
• Has a written action log completed before anyone leaves the room

A good one-page summary:

• Can be read in under two minutes
• Contains at least one specific example for the performance and trajectory assessments
• Does not use jargon or level-specific euphemisms ("they're punching above their weight")
• Is honest about what you do not know as well as what you do

Common Failures

The review becomes a performance management session for one person. If one individual takes up 40 minutes, the rest of the team gets shortchanged. Set a timer. If a person needs more discussion, flag it and return to them at the end.

Managers have not prepared. They arrive without summaries and speak from memory. Their assessments are surface-level and hard to challenge. Fix this: make preparation a hard requirement. No summary, no floor time for that person.

The facilitator does not challenge vague assessments. "They're great" goes unchallenged. "They're a bit inconsistent" goes unchallenged. Without challenge, the session produces the same picture the manager already had - no added value.

Actions are vague or unowned. "We should keep an eye on this" is not an action. Every output needs a person, a specific step, and a date.

Confidentiality is breached. A manager goes back to their team and mentions that someone was discussed as a flight risk. The individual finds out. Trust collapses. Be explicit about confidentiality before you start and hold managers to it.

The 9-box becomes the output. People get labelled and the label follows them. The 9-box is a conversation tool. The output is actions, not labels.

The same people dominate every review. High performers and visible individuals get most of the airtime. Quieter performers, people on leave, people who are solid but not flashy - they get two sentences. Use the roster as a forcing function.

The session runs out of time. You cover the first half of the team well and rush through the second half. Time-box every person strictly. The person who is tenth on the list deserves the same quality of conversation as the person who is first.

Checklist

Two weeks before

• Confirm attendees - manager, skip-level, HR partner
• Set the date and book the room or video call (90-120 minutes for a team of 8-12)
• Share the one-page summary template with all managers
• Confirm the roster - who is in scope for this review
• Assign each manager their list of people to prepare summaries for

One week before

• Chase managers for summaries - they are due 48 hours before the session
• Review summaries as facilitator - flag any that are too thin or missing key fields
• Prepare the action log template
• Confirm running order and time per person

48 hours before

• Summaries received from all managers
• Running order confirmed and shared
• Room or call booked and confirmed
• Ground rules prepared

Day of session

• Ground rules read aloud at the start
• Timer running for each person
• Action log being completed in real time
• Facilitator challenging vague assessments
• Disagreements named and resolved or escalated

End of session

• Every person has a named action with an owner and a due date
• Action log read back before the group disperses
• "Who did we miss?" question asked
• Confidentiality reminder given
• Action log distributed to managers within 24 hours

Four weeks after

• Actions reviewed for completion
• Any actions not taken followed up
• Retention risk conversations that were agreed - have they happened?
• Note what to do differently next time

Reference: One-Page Summary Template

Name:
Role / Level:
Manager:
Time in role:

Current performance:
[One paragraph. What are they delivering? Against what expectations?]

Trajectory:
[ ] Growing  [ ] Stable  [ ] Declining
Why: [One sentence. What specifically is driving this assessment?]

Retention risk:
[ ] Low  [ ] Medium  [ ] High
Primary driver: [Compensation / Growth / Manager / Personal / External]
What you are seeing: [One sentence]

Development need:
[One specific, behavioural development need. Not a personality trait.]

Next role:
[What is the likely next step? Timeframe?]

What you do not know:
[What would you need to know more about to have a better picture?]

Reference: Example Conversation Fragments

Opening a challenge:

"When you say trajectory is high - can you give me a specific example from the last quarter of something that tells you that?"

Handling a vague retention risk:

"You've flagged retention risk as high. What are you specifically seeing? Have they said something? Changed behaviour? Had offers?"

Surfacing a gap:

"You've listed development need as communication. Can you be more specific - communication in what context, with whom, at what level of abstraction?"

Handling disagreement:

"So you're saying high potential and [other manager] is saying medium. Let's not average this. What is each of you seeing? Talk to each other, not to me."

Closing an action:

"So the action is: [manager name] has a conversation with [person] about the platform team role before end of March. Is that right? Is that achievable?"

When no action is proposed:

"We've talked about this person for five minutes and we haven't agreed an action. What do they need from us that they are not getting? There must be something."

Reference: Frequently Asked Questions

Should we share the talent review output with individuals?

No - not directly. The output is actions taken by managers, not labels given to individuals. You do not tell someone they are "high potential" or "retention risk high." You have a conversation with them about their development and what you are doing to support it. The action flows from the review. The label stays in the room.

What if a manager disagrees with the final agreed action?

They raise it in the session - not after. The session is the place for disagreement. If they feel strongly enough that an agreed action is wrong, they escalate to the skip-level before taking action. They do not ignore the output.

How often should talent reviews happen?

Twice a year is the minimum. Once if the team is very stable and small. More frequently - quarterly - if the team is growing fast, attrition is high, or the organisation is going through significant change.

What happens if someone is not included in the review?

That is a gap and it needs to be corrected. No one should be invisible in a talent review. If someone was missed - a person on leave, someone who recently joined, someone whose manager forgot to include them - add them to the follow-up list and complete their review within two weeks.

What if the manager has not built enough of a relationship to assess someone?

That is itself an output. "I do not know this person well enough to assess them accurately" is a valid finding - and an action: the manager needs to invest in that relationship. It is not an excuse to skip the review.

Can peers or the individual themselves contribute to the review?

Yes - as inputs, not as participants. A manager can bring peer feedback or a self-assessment as evidence to inform their summary. The individual does not attend the session itself. The session is specifically a manager-to-manager conversation about the picture of the team.

What do we do with the one-page summaries after the session?

They are stored securely by HR and the manager. They are reviewed at the next talent review cycle to track progress against actions. They are not shared outside the people management chain.