In November, Priya's manager told her, in a genuinely warm one-on-one, that she'd had a strong half: she'd shipped the migration ahead of schedule, mentored two junior analysts, and fixed a reporting bug that had been quietly costing the finance team hours every month. He used the word "exceeds." Six weeks later, her written review said "meets expectations," with a single line added by HR: "rating adjusted during calibration." Nobody in that calibration meeting had watched her work, read her code, or sat in on her stakeholder conversations. They had a spreadsheet, a curve, and forty-five minutes. Priya isn't unusual — she's typical of what happens at any company large enough to run a formal calibration process, which in practice means most employers with more than a couple hundred people on the payroll.
That gap between the spoken review and the written one isn't a paperwork glitch. It's the whole system working as designed. Most companies tie annual bonus pools, merit-increase budgets, and promotion slates directly to the rating that comes out of calibration, not the one your manager mentioned in your one-on-one — which is exactly why understanding the mechanics of that meeting matters as much as doing the work itself.
What Actually Happens Once You Leave the Room
Once a company grows past a few hundred employees, a single manager rarely gets to just decide your rating and move on. Instead, that manager sits down with five, ten, sometimes twenty peer managers — often with an HR business partner running the meeting — and works through every name on a shared spreadsheet, department by department. This is calibration, and depending on the company, it might also get called leveling, norming, or, in its older and blunter form, stack ranking.
The idea traces back further than most people assume. General Electric ran a formal version of this for two decades under Jack Welch, who called it the Vitality Curve: managers ranked their teams every year, and the bottom 10 percent were, by policy, let go. Microsoft ran its own version — internally nicknamed "stack ranking" — well into the 2010s, forcing managers to slot a fixed share of their team into lower buckets regardless of how the team performed as a group; the company only retired it in 2013, after years of public reporting on how it pitted high performers against each other and discouraged collaboration. Most companies today have softened the mechanics, and few still mandate a hard bottom-10-percent cut, but the underlying structure survives almost everywhere: a fixed or "guided" distribution of ratings, decided by a group, in a meeting the employee never attends. Whatever the company calls it on its careers page, the test is simple — if your rating has ever moved between what your manager said in the room and what showed up in the system three weeks later, you've already been through a calibration process, whether anyone used that word with you or not.
Why "Meets Expectations" Out Loud Can Become Something Else on Paper
The room runs on trades.
Say a department has thirty people and the guidance allows only four "exceeds" ratings this cycle. Eight managers walk in, each with at least one report they genuinely believe deserves that top slot. Somebody has to lose, and the outcome usually depends less on the work itself than on how much capital each manager has banked — whether they spent a top rating on someone else last cycle, how senior they are relative to the rest of the room, and how hard they're willing to push before it turns personal. A manager who gave you glowing verbal feedback in your one-on-one may walk into calibration fully intending to fight for you, get outmaneuvered by someone louder or higher up the ladder, and come out with a rating one notch below what they promised — without ever telling you the fight happened at all.
The Data That Survives Calibration — and the Data That Doesn't
A manager sitting in that room isn't reading your self-assessment out loud. They're making a case to peers who don't know your daily work and have maybe thirty seconds of attention to spare on any one name. What actually moves a rating in that room is evidence that didn't originate from your manager alone, because a manager vouching for their own report is the weakest form of proof calibration recognizes. Here's what tends to carry weight instead:
- A specific number tied to a business outcome — revenue protected, hours saved, a deadline hit that others missed — rather than a general description of effort.
- A quote or comment from someone outside your reporting line, especially if that person happens to be in the room or known to the people who are.
- Visibility with your manager's manager, since a skip-level who already has an opinion of your work can speak up during the trading instead of your manager fighting alone.
- Written praise with a name and date attached, like a Slack message, an email, or a project retro your manager can point to rather than paraphrase from memory.
Vague effort — the fact that you stayed late, took on extra meetings, or were generally reliable — rarely survives contact with that room, because none of it translates into a sentence a manager can say out loud to people who have never worked with you.
Three Moves to Make Before Your Manager Walks Into That Room
You can't sit in on calibration, and asking to would be a strange, career-limiting request in almost every company culture. What you can do is change what your manager carries into that room before the door closes — and the moves that actually work happen weeks before the meeting, not after you've already seen a disappointing number on paper.
Ask the Blunt Question Before the Window Closes
Most review cycles have a window, usually two to four weeks, between when self-assessments are due and when calibration happens. That window is your last real chance to affect the outcome, and almost nobody uses it. Book ten minutes with your manager and ask directly: "Where do you think my rating lands right now, and is there anything you need from me before you go into calibration?" Ask it plainly — don't soften it into "how do you think things are going," because that invites a comfortable, vague answer instead of the specific bracket you actually need. A manager who has to answer a direct question out loud will often tell you things they'd never volunteer unprompted, including which peer manager they expect to clash with over a contested slot.
Feed the Room Secondhand, Not Just Firsthand
Since a manager vouching alone is weak evidence, the more useful move is making sure other names surface for you before the meeting, not after. If you finished a project with a stakeholder in another department, ask that person — while the memory is still fresh, not months later — to send a short note to your manager describing the specific impact, two or three sentences with a real number in them. Do this two or three times a cycle, not once. One secondhand mention reads as a nice compliment; three, from three different people, becomes something your manager can actually quote in the room without sounding like they're just repeating their own opinion.
Get a Skip-Level Meeting on the Calendar Mid-Cycle
Request time with your manager's manager at the midpoint of the review period, not at the end, when it reads as a last-minute pitch. Use it to talk through one or two concrete pieces of work rather than a summary of your whole role. The goal isn't to go over your manager's head — it's to make sure the person sitting in calibration with an opinion about your work formed that opinion firsthand instead of secondhand from a spreadsheet cell. A skip-level who remembers a specific conversation with you is far more likely to speak up when your name comes up for a contested slot than one who only recognizes it from an org chart.
When the System Is Broken, Not You
None of this guarantees a better outcome, and it's worth saying plainly: sometimes the math is stacked against you no matter what you do. If you land on a team that's unusually strong this cycle, a guided distribution can force your manager to rate someone below their actual performance simply because too many people on that team are competing for too few top slots — and if a recent layoff or reorg squeezed your department's budget for "exceeds" ratings this year, that scarcity has nothing to do with your work at all. A pre-calibration check-in, secondhand praise, and a skip-level relationship all improve your odds; they don't override a curve that was set before the cycle even started. If you do everything right and still come out with a flat rating two cycles in a row, treat that as a signal about the team or the company, not about you — that's the better read, not one more thing you should have documented. Sometimes the honest next move isn't a better pre-calibration pitch; it's asking your manager, plainly, whether the team's rating pool has been shrinking, and treating a shrinking pool as information about where to spend your energy next.
The calibration meeting will keep happening with or without you in the loop. The only real choice is whether your manager walks in with three sentences they can say with confidence, or with nothing more than a gut feeling and your name on a list.