Guide · Pay reviews

What performance calibration can't fix

Calibration fixes ratings. It does not fix budgets, ranges or the step where ratings become raises.

Short answer: Performance calibration makes ratings more consistent across managers. That is all it does. It cannot fix a team whose budget was too small, people paid low in their range or bias in the step where a rating becomes a raise. Those need their own checks. Bending ratings to fit the budget breaks the one thing calibration can fix.

A calm meeting and an angry week

Picture a 240-person company in Reno that makes warehouse software for online retailers. This year, for the first time, its leaders held calibration sessions before the pay review. Eight managers walked through their ratings together. Two who had rated nearly everyone "exceeds" moved several people down. One strict manager moved two people up. The meeting ran on time and everyone agreed the ratings now looked fair.

A week after the raises went out, the complaints started. The support team, rated fairly after calibration, got smaller raises than the sales operations team with similar ratings, because its budget pool had been set as a share of a lower payroll. Three engineers rated "exceeds" got less than colleagues rated "meets", because they already sat high in their ranges. And when the head of people looked at the final raises by gender, women with the same rating as men had received slightly less on average. Nobody had looked at that before letters went out.

The leadership team had expected calibration to make the pay review fair. It made the ratings fair. Everything that happened after the ratings was untouched.

This is a composite of a common situation at companies this size, not a client. The details and figures are illustrative.

Why calibration gets asked to do too much

Calibration sets two good goals against each other. The first is consistency. Managers rate differently and a meeting where they compare and adjust brings ratings closer to one standard. The second is fairness in pay, which is what employees actually experience. Leaders reasonably assume that consistent ratings lead to fair raises. They often do not, because a raise depends on several things besides the rating.

The best field evidence on calibration comes from a study of a large multinational published in 2019. About one rating in four was adjusted by its committees, with cuts more frequent and larger than increases. Ratings ended up closer from one supervisor to the next and less generous, but also more bunched toward the middle [1]. That is what calibration is good at: Narrowing the gap between generous and strict managers. It says nothing about budgets or about what happens next.

What happens next is where many pay gaps arise. A well-known study of merit pay at a large company found that women and minorities received smaller raises than white men with the same performance score, job, unit and manager. A later study at a company that added accountability and transparency to the step from rating to raise found the gaps narrowed [2]. The rating can be perfectly calibrated and the raise still uneven.

Managers add to the confusion without meaning to. Before calibration, a manager who rates generously is often protecting the team's raises as much as describing its performance. When calibration lowers those ratings, the manager experiences it as a cut to pay, not a correction to a rating. That is a sign the two have been fused in people's minds. The more a company relies on calibration to deliver fair pay, the more its managers will treat the meeting as a negotiation over money.

Ratings also carry forward. A field study at a multinational financial services firm found that managers anchored on the previous year's ratings, which limited the effect of changing how reviews were run [3]. A rating bent to fit this year's budget becomes the starting point for next year's.

Two terms will help. Calibration is a meeting where managers and leaders compare proposed ratings across teams and adjust them toward a shared standard. Position in range is where a person's pay sits between the minimum and maximum for their job.

Five traps when calibration meets pay

"Calibrated ratings mean fair raises"

What it looks like: The company invests its effort in the calibration meeting. Once ratings are agreed, raises follow from a merit guide or from managers' proposals with no second look.

Why smart teams fall into it: The calibration meeting is visible, tiring and feels like the fairness step. After two hours of debating ratings, a second review of the numbers feels redundant.

What it costs: The step from rating to raise is where judgment returns. Managers choose within ranges, apply discretion at the edges of the guide and handle people above or below their range. The research is clear that this is where gaps between groups arise and where accountability narrows them [2]. Without a look at the proposed raises by rating, team and gender before letters go out, calibration gives the company confidence it has not earned. Our guide on merit matrices and manager judgment covers where each approach breaks in the pay step.

"Calibration will fix the underfunded team"

What it looks like: Budgets are handed to teams as a share of their payroll. Calibration then makes ratings consistent across teams. Leaders expect raises to come out consistent too.

Why smart teams fall into it: Pools by team are simple to run and make each manager responsible for a number. The mismatch only shows when outcomes are compared side by side.

What it costs: A strong performer in a team with a small pool gets less than the same rating in a team with a larger one. Calibration cannot see this, because it only compares ratings. The fix is in how the budget is set: Pools that follow the people who need them, or a central reserve for teams whose people sit low in their ranges. A budget problem needs a budget answer.

Position in range is a second blind spot. Most merit guides give smaller raises to people already high in their range, which is a sound rule when the range reflects the market. The three engineers in the opening got less than colleagues with lower ratings because of that rule, not because calibration failed. But nobody in the calibration meeting could see it, so nobody prepared their managers to explain it. If range position drives the raise, it should sit next to the rating in every review of the numbers. The managers of people high in their range should know the reason before the conversation. Our guide on splitting the salary budget covers why market and structure money should not come out of the merit pool.

"Fit the ratings to the budget"

What it looks like: When the proposed raises cost more than the budget, the calibration session lowers ratings until the numbers fit. The conversation is about money, presented as a conversation about performance.

Why smart teams fall into it: Lowering ratings is the fastest way to bring a merit guide back inside the budget. It avoids reopening the guide or asking for more money.

What it costs: The ratings stop meaning what they say. Managers learn that ratings are a budget tool and start to rate defensively. Employees whose ratings fall without a change in their work lose trust in the whole system. The lowered ratings also last. They anchor next year's ratings [3] and they become the record the company would rely on to defend pay differences. Under US federal guidance, a merit system is a defense only if its criteria are set in advance and applied consistently [6]. Ratings adjusted to hit a cost target are hard to describe that way. If the budget is too small, change the merit guide or the budget, not the ratings.

"A curve keeps the raises honest"

What it looks like: To control inflation and cost, the company sets a target distribution of ratings, such as 10% at the top and 20% in the next band. Calibration then forces each team into it.

Why smart teams fall into it: Rating inflation is real. The US federal personnel office reported that in 2024 nearly 43% of employees below senior levels on a five-level scale got the top rating and it has issued target distributions for the 2026 and 2027 cycles [4]. A curve looks like a cure.

What it costs: At 240 people, most teams are too small for a curve to describe them. A team of six strong engineers forced to rate one as below expectations is not more accurate. It is less. A curve also adds to the pull toward the middle that calibration already creates [1], which narrows the difference in raises between strong and average performers, the opposite of what most companies want. Guidance on what a typical distribution looks like can help. A quota applied to every team usually does not.

"Calibrate after the numbers are drafted"

What it looks like: Managers propose ratings and raises together. Calibration happens at the end, when the raises are already in a spreadsheet and sometimes already hinted at to employees.

Why smart teams fall into it: Timelines are tight and it seems efficient to review everything in one meeting.

What it costs: Once a number is attached, the discussion about the rating becomes a discussion about the money. Managers defend ratings to protect raises they have already promised. Changes made in the meeting then have to be walked back with employees, which is the worst way to deliver a rating. Calibration works best on ratings alone, before anyone has seen a raise. Pay comes after and gets its own review.

What calibration that helps pay looks like

Good calibration does not try to fix the pay review. It does one job well and leaves the other jobs to the checks built for them. At 50 to 500 people, the processes that hold up share these qualities.

Ratings are calibrated before raises exist. The meeting discusses performance against written criteria. Nobody in the room is defending a number.

The meeting focuses on the edges. Time goes to the highest and lowest ratings and to managers whose ratings differ most from the rest. Not every rating needs debate.

The data is in the room. Each manager's distribution of ratings, last year's ratings and the split by gender and team are visible, so patterns are noticed while they can still change.

The budget is set separately. Money for market gaps, promotions and people low in their range is decided outside calibration and outside the merit pool, so the ratings never have to absorb a budget problem.

Range position sits next to the rating. When proposed raises are reviewed, each person's position in range is visible beside their rating, so a small raise for a strong rating is explained by the rule that produced it, not discovered by the employee.

The raises get a second look. Before letters go out, someone reviews proposed raises by rating, position in range, team and gender. That is where calibration's blind spot is covered.

Distribution guidance, not quotas. If leaders share what a typical spread looks like, managers who differ sharply explain why. Small teams are not forced into a curve.

Decisions are recorded. Each change of a rating has a short reason. That record is what lets the company explain a rating later, to the employee or anyone else.

Managers hear the outcome before their conversations. Our guide on explaining a small raise to a strong performer covers what to say when a calibrated rating still leads to a small number.

The calls only you can make

Target distributions. Guidance curbs inflation and adds pressure toward the middle. A quota does both more strongly. Decide which you want and say so.

Who sits in the room. Managers only, managers with their own leaders, or a small panel with someone from HR. A broader room catches more and moves slower.

Whether employees see their rating. Some companies share ratings, others share only the raise. Sharing makes calibration matter more to employees and makes changes harder to explain.

The power of the room over a manager. A committee that can override a manager's rating creates consistency and can create resentment. One that can only ask questions leaves more inconsistency in place.

What money sits outside the merit pool. Market adjustments, promotions and range fixes can be funded separately or not at all. Whatever you decide, keep it out of the rating discussion.

The research and the numbers

  • What calibration changes. A study of more than 1,300 performance evaluations at a large multinational found that calibration committees adjusted about one rating in four, with cuts more frequent and larger than increases. Ratings ended up closer from one supervisor to the next and less generous, but more bunched toward the middle [1]. This study is older than the others here. We cite it because it is the main field study of how calibration committees behave.
  • Where pay gaps arise. Research on merit pay at a large company found smaller raises for women and minorities with the same performance score, job, unit and manager as white men. When accountability and transparency were added to the step from rating to raise, the gaps narrowed [2].
  • Ratings carry forward. A field study at a multinational financial services firm from 2015 to 2018 found race gaps in manager ratings and found that managers anchored on the previous year's ratings, which limited the effect of changing the review process [3].
  • Rating inflation. In 2024, nearly 43% of US federal employees below senior levels on a five-level scale received the top rating. The federal personnel office has issued example targets of about 10% at the top level and 20% at the next for the 2026 and 2027 cycles [4].
  • Raises by rating. A compensation software provider's data for the first half of 2026 shows median raises of 4.7% for people rated above expectations, 3.3% for those meeting them and 2.0% for those below [5]. These are the provider's own figures.
  • United States, merit systems. Under federal equal pay guidance, a merit system is a defense only if it was not adopted with discriminatory intent, is based on criteria set in advance, has been communicated to employees and is applied consistently. Systems that rely heavily on subjective judgment get close scrutiny [6].
  • European Union. The pay transparency directive requires the criteria used for pay progression to be easily accessible to workers, objective and gender-neutral (Article 6) [7].

This is general information, not legal advice. Rules and their national versions differ. Check the rules where you employ people.

Our view

We think calibration is worth doing and worth keeping narrow. It is the best tool a company has for bringing generous and strict managers onto one standard. It is the wrong tool for a small budget, a team paid low in its ranges or bias in how ratings turn into raises.

So we would calibrate ratings before any raise is drafted, fund market and structure gaps outside the merit pool and review the proposed raises by group before letters go out. And we would never change a rating to make the numbers fit. Calibration earns its keep when everyone in the room knows it is about ratings and that the money will be checked separately. A company that keeps its ratings honest can fix its pay. One that bends its ratings has lost the tool it needs to do it.

Sources

  1. Demeré, Sedatole and Woods, The role of calibration committees in subjective performance evaluation systems, Management Science 65(4), 2019, pp. 1562 to 1585, checked 2026-10-30
  2. Castilla, Accounting for the gap: A firm study manipulating organizational accountability and transparency in pay decisions, Organization Science 26(2), 2015, checked 2026-10-25
  3. Bohnet, Hauser and Kristal, Can gender and race dynamics in performance appraisals be disrupted? The case of social influence, Journal of Economic Behavior and Organization 235 (2025), summary by the Harvard Kennedy School, checked 2026-10-29
  4. FedWeek, Forced distributions will apply to current ratings cycle, says OPM (18 August 2026), checked 2026-10-30
  5. Pave, merit increase data for the first half of 2026 (26 June 2026), checked 2026-10-30
  6. US EEOC, Compliance Manual Section 10: Compensation Discrimination, 10-IV.F.1 (merit and incentive systems), checked 2026-10-29
  7. Directive (EU) 2023/970 on pay transparency, Article 6 (EUR-Lex), checked 2026-10-28

Merit matrix or manager judgment?

Merit matrices fail at the edges and manager judgment fails at scale. Five ways merit decisions break and what a sound process looks like.

How should you split the salary budget?

When one budget pays for promotions, market fixes and range moves, merit gets what is left. Five traps and what a sound budget split looks like.

Your first performance and pay cycle

A first pay cycle sets habits for years. Five traps when a startup runs its first one and why one clear decision per person beats a big process.

See Resources for every guide and the self-check.