> ## Content Index
> Fetch the complete content index at: https://proofweight.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Do ratings actually improve performance?
- URL: https://proofweight.com/do-performance-ratings-improve-performance/
- Published: 2026-09-09T06:44:26.000Z
- Updated: 2026-09-22T11:25:07.000Z
- Description: Rarely on their own. A rating is a decision aid, and it serves whoever it was designed for. Decide who that is before you defend it or dismantle it.
- Author: Aditya Rao
- Tags: Performance, Performance management, #status-contested, #has-sources

**Rarely on their own. A rating is a decision aid, and it serves whoever it was designed for. Decide who that is before you defend it or dismantle it.**

*For the CHRO, HR leader or business leader who owns the performance system. Article 4 of seven in the performance management investigation.*

A manager tells someone they had a strong year, and that the way they worked with operations made a real difference. Then the review opens: meets expectations. The manager explains that it went through calibration and the reward budget was tight.

The employee has one reasonable question, and it is the question this article is about. Which part of that rating describes my work?

## Ask who the rating is for

Before asking whether ratings work, ask what they are for. The answer is usually four different things at once, and they pull in different directions.

| Who                   | What a rating gives them                             | What they lose without it                              |
| --------------------- | ---------------------------------------------------- | ------------------------------------------------------ |
| The employee          | A statement of where they stand, and what was valued | Clarity, and a thing they can argue with               |
| The manager           | A way to summarize and communicate contribution      | A shorthand, and cover for a hard conversation         |
| HR and process owners | Comparability, consistency, an auditable record      | Control over how judgments are made across the company |
| The leadership team   | A common currency for pay, promotion and succession  | The ability to compare people they have never met      |

That multiplicity is not a new observation. It has been studied directly [1](#sources): organizations run these systems for several purposes at once, and the purposes shape how the system behaves.

**As practitioners we have observed**, and this is the practical point: a system designed for the fourth row will not serve the first. If leaders need comparability across two thousand people, the number will be built for them, and the employee will get a summary of a negotiation they were not in.

## The version everyone remembers

The best known rating system of the last thirty years is the one General Electric ran under Jack Welch, sorting people into roughly a top twenty percent, a vital seventy, and a bottom ten, with consequences attached to the bottom.

What did it solve for? Not employee clarity. It was built to force differentiation where managers avoided it, and to move people out.

There is evidence on that specific design. A simulation study of forced distribution systems found they could produce a noticeable improvement in average workforce potential, that most of the improvement should be expected over the first several years, and that the size of it depends mainly on the percentage fired and on how many people leave voluntarily anyway [2](#sources).

Read that carefully, because it is often quoted as proof that ratings improve performance. It is a model of improving the workforce by replacing people, not of the people you keep getting better. It also predicts the thing practitioners observed: the gains fade, because after a few cycles you are firing people who are not obviously worse than their replacements.

## So do ratings improve performance?

The honest answer is that this is still the open question, a century into the literature. The field's own centenary review is titled "Performance appraisal and performance management: 100 years of progress?" with the question mark included [3](#sources).

Two more things are established and worth separating from the argument. Training managers against shared standards can make judgments more accurate [4](#sources), which is a claim about the quality of the judgment rather than about the organization's results. And the published debate on abolishing ratings sets out serious arguments on both sides, without an experiment to settle it [5](#sources).

**As practitioners we have observed:** if the decision you are making is administrative, a rating can be defensible and efficient. Someone has to allocate a pay pool and decide who is ready for more. If the aim is improving performance or engagement, the rating is a weak lever, and the evidence that it moves either is thin.

## The pay question, which is the real one

In many organizations, pay is the reason ratings remain. The rating is the instrument that converts a year of work into a number that can be multiplied by a budget.

That linkage has consequences, and they are worth naming before you decide whether to keep it.

**The rating stops being about development.** Once a number sets an increment, nobody in the conversation is optimizing for learning. The manager is defending an allocation and the employee is protecting an outcome, and the developmental conversation the system claims to run quietly stops happening.

**Distributions get imported from the budget.** When the pay pool assumes a spread, the ratings will produce one, whatever the work looked like.

**Ratings compress.** Where managers must justify low ratings and fund high ones, most people end up in the middle, which is the least informative place on the scale.

**The judgment becomes contested rather than useful.** A number attached to money invites negotiation, and negotiation rewards the confident.

Now the contradicting evidence, because it matters here. The intuitive case, that tying pay to performance corrupts motivation, is weaker than it sounds. A meta-analytic review examined the relationship between financial incentives and performance [7](#sources), and a later review argued directly that the common cautions about pay harming intrinsic motivation and creativity lack support in workplace settings [8](#sources).

**As practitioners we have observed** that both things hold at once. Pay linkage does not destroy motivation, and it does damage the rating as a description of work. So the practical question is not whether to pay for performance. It is whether the same instrument should do both jobs.

The alternative is unfashionable and simple: make the performance judgment first, record the reasons, and let the reward decision be a separate step with its own logic, its own constraints and its own explanation. You still pay differently. You stop pretending the difference was purely a judgment about the person.

## When the budget rewrites the year

One failure mode is worth naming on its own, because it does more damage to trust than any scale design.

Suppose a rating is lowered because too many people qualify for the top reward. Either the initial ratings were generous and the discussion corrected them, or the assessments were sound and the money was short. The first is a judgment. The second is a constraint.

Calling both calibration hides which problem the organization has, and leaves the employee with an account of their year that quietly changed to fit a budget. Recording the performance judgment before the reward allocation, and the reasons for the allocation separately, does not make the decision easier. It makes it inspectable, and inspectable is closer to what the research on fair process says people actually respond to [6](#sources).

## What the evidence supports, and how far

Appraisal serves several purposes at once, and that has been studied rather than assumed [1](#sources). Forced distribution can raise average workforce potential in a simulation, mostly in the early years, driven by how many people are removed [2](#sources). A century on, the field's own review still frames progress as a question [3](#sources). Shared-standard training improves rating accuracy [4](#sources). The debate on removing ratings is a debate, not an experiment [5](#sources). How a decision is made, and whether it can be explained, has a large meta-analysed literature of its own [6](#sources).

**The stakeholder table and the separation of judgment from reward allocation are ours.** Nothing here establishes that an annual rating improves the performance of the people who receive it.

**Evidence: contested.**

## Our position

**Keep a rating if you need to allocate pay and compare people across a large organization, and stop claiming it improves performance. Separate the performance judgment from the reward decision, and be able to explain both.**

**Take it into the room:** the [Put the rating to work](https://proofweight.com/worksheets/?template=test) worksheet turns this article into one printed page of questions, a framework and a table to fill in.

## What would change this conclusion

A comparison in which the rating element changed and everything else held steady, tracking the decisions and actions that followed. Or evidence that employees who receive ratings understand their priorities better than those who do not, which is the claim most rating systems implicitly make and almost none test.

**Read next: [Does calibration make performance decisions fairer?](https://proofweight.com/does-calibration-make-performance-decisions-fairer/)**

## Evidence map

1. **Cleveland, Murphy and Williams (1989), [Multiple uses of performance appraisal: Prevalence and correlates](https://doi.org/10.1037/0021-9010.74.1.130?ref=proofweight.com).** Journal of Applied Psychology. Record and title verified through Crossref and Semantic Scholar; abstract not retrievable through the sources checked. Cited for the line of inquiry its title establishes.
2. **Scullen, Bergey and Aiman-Smith (2005), [Forced distribution rating systems and the improvement of workforce potential](https://doi.org/10.1111/j.1744-6570.2005.00361.x?ref=proofweight.com).** Personnel Psychology. Published abstract accessed in full. A simulation of workforce composition through replacement, not a study of individual improvement.
3. **DeNisi and Murphy (2017), [Performance appraisal and performance management: 100 years of progress?](https://doi.org/10.1037/apl0000085?ref=proofweight.com)** Journal of Applied Psychology. Record and title verified; abstract not retrievable. Cited for what the title establishes about the state of the field.
4. **Roch, Woehr, Mishra and Kieszczynska (2012), [Rater Training Revisited](https://doi.org/10.1111/j.2044-8325.2011.02045.x?ref=proofweight.com).** Meta-analysis; publisher abstract accessed.
5. **Adler and colleagues (2016), [Getting Rid of Performance Ratings: Genius or Folly? A Debate](https://doi.org/10.1017/iop.2015.106?ref=proofweight.com).** Published debate; relevant full text accessed.
6. **Colquitt, Conlon, Wesson and colleagues (2001), [Justice at the millennium: A meta-analytic review of 25 years of organizational justice research](https://doi.org/10.1037/0021-9010.86.3.425?ref=proofweight.com).** Journal of Applied Psychology. Record and title verified; abstract not retrievable through the sources checked. Cited for the existence and scale of the literature on how fair process affects reactions, not for an effect size.
7. **Jenkins, Mitra, Gupta and Shaw (1998), [Are financial incentives related to performance? A meta-analytic review of empirical research](https://doi.org/10.1037/0021-9010.83.5.777?ref=proofweight.com).** Journal of Applied Psychology. Record and title verified through Crossref; abstract not retrievable through the sources checked. Cited for the line of inquiry its title establishes.
8. **Gerhart and Fang (2015), [Pay, Intrinsic Motivation, Extrinsic Motivation, Performance, and Creativity in the Workplace: Revisiting Long-Held Beliefs](https://doi.org/10.1146/annurev-orgpsych-032414-111418?ref=proofweight.com).** Annual Review of Organizational Psychology and Organizational Behavior. Published abstract accessed in full. The authors assess the evidence behind the usual cautions about pay for performance and conclude that support for those claims in workplace settings is lacking. **This contradicts the intuition many practitioners hold, including ours, and we include it for that reason.**
9. **General Electric's differentiation system** is described from its widely reported design, not from a study.

Sources checked September 22, 2026\. Prepared with AI assistance.