A calibration session compares judgements; it cannot improve the evidence those judgements were made from.
Why it matters when the plan changes
Discussing people across a leadership group surfaces disagreement that individual reviews hide, and reduces the influence of one manager's view. That is genuinely valuable. What the format cannot do is manufacture information, so a room full of confident impressions produces a confidently calibrated set of impressions. Information also weakens as it travels through a reporting line, so what each manager brings into the room is already partial before the discussion starts, and calibration cannot restore what was lost on the way up.
The tension is that the session is where organisational politics meets individual futures. Advocacy skill, visibility and the relative standing of the managers present all bear on the outcome, and none of them is about the person being discussed. Confirmation bias makes this worse: people search for and recall information that supports a view they already hold, which is exactly what an evidence-free discussion invites. Structure and evidence are the only counterweights, and both have to be brought in rather than generated on the day.
In practice
A leadership group spends a day placing names on a grid. The placements shift according to who argues most effectively for their own people. The output is recorded as a calibrated view and is used for two years of development decisions, and no record exists of what evidence supported any individual placement.
Evidence
Information about people weakens as it travels through a reporting line, so what each manager brings into a review is systematically partial.
Donald Sull, Rebecca Homkes and Charles Sull, Why Strategy Execution Unravels and What to Do About It, Harvard Business Review (2015)People search for, favour and recall information that confirms their existing view, which is what a discussion without evidence amplifies.
Confirmation bias, Wikipedia (2026)
What it cannot tell you
A talent review can surface disagreement between managers and produce a consistent group view, but it cannot tell you whether that view is correct. It has no mechanism for checking the evidence behind a judgement; if managers arrive with impression rather than observed behaviour, the session amplifies agreement, not accuracy.
Questions
A shared view of who is performing, who is ready for more, who is at risk of leaving and where the succession gaps sit, often placed on a grid. The output feeds development budgets, succession plans and sometimes pay, which is why the quality of the inputs matters.
It improves consistency between managers, not accuracy. Sull, Homkes and Sull, writing in Harvard Business Review in 2015, found that information degrades as it passes up a reporting line, so the judgements entering the room are already partial. Calibration cannot correct that; it only makes the partial judgements agree with each other.
Advocacy skill, the relative standing of the managers present, recency effects, and how visible individuals happen to be. A 2026 Wikipedia entry on confirmation bias describes the tendency to search for and recall information that confirms a view already held, which an evidence-free discussion invites. Requiring written evidence for each claim is the main counterweight available.
Specific and written down before the session: observed behaviour against the demands of the role, with whatever argues against the judgement recorded alongside it. Evidence assembled during the discussion itself will be selected to support the position somebody is already arguing for.
It should be anchored to one. Reviewing people against a general standard produces a ranking that may not bear on what the organisation is about to attempt. Reviewing them against what the plan will demand makes the session decide something specific.