The Calibration Meeting: The Closed-Door Session That Actually Decides Your Performance Rating

Your self-assessment is due on Friday. It will take you two hours, three drafts, and a fair amount of second-guessing over whether "consistently exceeded expectations" sounds arrogant. None of that effort decides your rating. What decides it happens the following week, in a room you're not in, where your manager sits across a table from six or eight other managers and argues your number up or down against people you've never met.

The Meeting You've Never Been Invited To

It's called a calibration meeting, sometimes a calibration committee, and most large employers run one before ratings are finalised and communicated to staff. The mechanics are simple enough to describe and almost never explained to the people they affect. A group of managers — usually peers within the same function or region — bring their proposed ratings for every person on their teams. One by one, the ratings go up on a screen or a whiteboard, and the group debates whether a "strong performer" in one manager's team would actually be a "strong performer" if judged by the standards applied two desks over. A manager who rates everyone highly gets pushed back. A manager who's harsh across the board gets the same treatment in reverse.

The stated purpose is fairness — correcting for the fact that individual managers have wildly different bars, some generous, some brutal, almost none of them calibrated against each other without outside pressure. In practice it also functions as a forced conversation about scarcity: most organisations don't have an unlimited pool of "exceeds expectations" ratings or promotion budget to hand out, so somebody's rating moves down so somebody else's can move up. Your manager can walk in wanting to rate you a 4 out of 5 and walk out having agreed to a 3, not because your work changed, but because the room needed one fewer 4 that cycle.

Where This Practice Came From

Calibration in its current, softer form is largely a reaction to a much harsher predecessor. General Electric ran what became known internally as the "vitality curve" from the early 1980s, under the CEO whose name became synonymous with forced ranking: employees sorted every year into a 20-70-10 split — 20% top performers, 70% solid middle, and a bottom 10% who were typically managed out regardless of whether their absolute performance was actually poor. Critics called it "rank and yank." GE itself abandoned the practice in 2015, by which point the model had spread to dozens of other large companies and drawn years of criticism for pitting colleagues against each other and punishing people simply unlucky enough to land on a strong team.

Microsoft ran its own version — an internal stack-ranking system that required managers to slot a set percentage of their team into lower tiers no matter how the team as a whole had performed. The company scrapped it in November 2013, with its HR chief announcing the change internally after years of employee complaints that stack ranking encouraged people to sabotage colleagues rather than collaborate with them. Google took a different route entirely: its former head of People Operations described, in a widely read book on the company's management practices, how Google's calibration sessions bring groups of managers together specifically to compare notes and flag outliers, without a forced curve dictating how many people land in each bucket. And Amazon's calibration process made headlines in 2021 when Business Insider reported on leaked internal documents describing an "unregretted attrition" target — reportedly around 6% of the workforce per year — that some teams used to identify who should be pushed toward a performance improvement plan. Amazon has publicly disputed the quota framing, telling reporters at the time that performance decisions were based on individual results rather than a fixed numerical target. Whichever version is closer to the truth on any given team, the reporting made one thing visible that most employees never see directly: a specific attrition or distribution number, discussed among managers, sitting upstream of the rating that eventually reaches an employee's inbox. The common thread across all four is the same: a rating your manager proposes is not the rating you receive until a group of people who don't work with you directly has had a say in it.

Why Mid-September Is When It Starts

For companies running a calendar fiscal year — the majority of large employers — September is when the machinery behind annual reviews starts moving, even though nothing visible reaches employees until October or November. Merit increase budgets get modelled against finance's Q4 forecasts around this time. HR business partners start scheduling calibration sessions for late September through October, well ahead of the December compensation cycle most companies run. If your manager hasn't yet had a serious conversation with you about how your year has gone, the window to influence what they walk into that room with is closing, not open indefinitely.

What Actually Moves the Outcome

You cannot sit in the calibration room.

You can absolutely change what your manager brings into it, and most people leave this to chance.

• Get your strongest, most specific evidence in front of your manager two to three weeks before their calibration date, not during the formal review window when everyone else is doing the same thing and their attention is split ten ways.
• Ask your manager plainly whether a calibration process exists on your team, and when it runs. Most managers will tell you if you ask directly — they're not sworn to secrecy, they simply assume nobody wants the mechanical detail.
• Understand where your manager sits relative to their peers. A manager who is new, unpopular with other leads, or manages a small team walks into calibration with less leverage than a manager who's respected and vocal in the room — and that gap affects you whether or not you ever meet the other managers.
• Push for specifics rather than adjectives in whatever your manager says about you before the meeting; "strong" and "solid" get argued down in calibration far more easily than a named deliverable with a date and a measurable result attached to it.

None of this guarantees the outcome you want. What it does is put your manager in the room with something concrete to defend, instead of a vague, favourable impression that dissolves the moment somebody else in the room pushes back.

The Limit of Your Influence

There's a version of this advice that oversells what preparation can do, and it's worth naming honestly. If your manager has no real standing in the calibration room — junior, disliked, or simply outnumbered by peers advocating harder for their own people — no amount of documentation you hand them will fully offset that structural disadvantage. Calibration is, in the end, a negotiation among managers, and you are dependent on someone else's negotiating position whether you like it or not. The honest response to that isn't to give up on preparation; it's to also pay attention to whether your manager has real credibility with their peers, because that's a signal about your own ceiling on that team that documentation alone can't fix. One rough proxy worth watching: whether your manager is the one presenting updates in cross-team meetings, or the one sitting quietly while someone else runs the room. It's not a perfect measure of standing, but it's a visible one, and it costs you nothing to pay attention to it over a few months.

Ask the Question Nobody Else Is Asking

Most employees never ask their manager the one question that would tell them the most: "Is there a calibration meeting before ratings are set, and roughly when does it happen?" It's a reasonable, professional question, not an accusation, and a manager who respects you will usually answer it straight. If the answer is yes, you now know exactly when your evidence needs to land. If the answer is no — some smaller companies and most startups skip formal calibration entirely — then your manager's word genuinely is closer to final, and the entire calculation around timing changes. Either way, you're no longer guessing about a process that was quietly shaping your outcome long before your self-assessment was ever due.