Jess Kovach

Self-Assessment Examples, and What They Can't Fix

How to write a self-assessment, with examples, what to do when the criteria are vague, and what the person reading it owes.

On this page
  1. What should you put in a self-assessment?
  2. How do you write about results nobody measured?
  3. What do you do with work the criterion doesn't count?
  4. What if the rating criteria are vague?
  5. Should you rate yourself higher?
  6. What if writing it is the hard part?
  7. Should companies stop asking for self-assessments?
  8. What does the person reading it owe?
  9. What a self-assessment can't do
A blank spiral notebook page and a red pen on a dark desk, seen from above.

The self-assessment is the one document in your review that you write. The rating, the comments, and what gets said about you in the calibration meeting all come from someone else.

It also gets read early. In a study of more than 100,000 performance reviews from 170 companies, 90 percent were set up so that managers could see the employee's other reviews, including the self-review, while writing their own (Abraham, 2023). The score you give yourself can pull on the score you get (Bohnet et al., 2025).

So write it well. Just don't expect it to do more than it can. The bias in a review often isn't on the form. It's in how the person writing your review applies the form: what they notice, what they leave out, what they tell you, and what they don't. A self-assessment can't remove that.

Here's what it can do. It can put your results on the record in the words of the standard. And it can show whether there was a standard at all. Setting one is the job of the person reading it, so this post is for them too.

What should you put in a self-assessment?

Your results, each one written against the criterion it meets, with the criterion quoted. Not adjectives about yourself.

Start with the criteria that were actually stated: the goals set at the start of the year, the job description, the competency model, the definitions on the rating scale. Copy the exact words. Then, for each one, write what you did, how it was measured, and who saw it.

A weak line describes you. A strong one describes the work, in the standard's own words.

Criterion: "Develops others to perform independently."

Weak: "I'm a team player who always goes above and beyond."

Written to the criterion: "Develops others to perform independently: I trained four new associates on returns between March and June. By July, all four were closing returns without a lead's sign-off."

Rae, the shift lead who wrote that line, also had a goal on paper: "Reduce open weekend shifts." Their first draft said, "Worked hard to improve scheduling." Written to the goal, it says: "Reduce open weekend shifts: open weekend shifts went from about six a month in January to one or two a month by June, after I moved shift swaps to a shared sign-up sheet. The store manager approved the sheet in February."

Dana is Deaf and works through an interpreter. Her competency model lists "Communicates effectively with stakeholders," and the easy line is "Strong communicator." Written to the criterion: "Communicates effectively with stakeholders: ran the weekly vendor call for the system change from April to September, through an interpreter, and sent written notes after every call. All 14 vendors were on the new system by the September deadline." The criterion asks whether the communication worked. It doesn't ask what it sounded like.

How do you write about results nobody measured?

A hand writing figures in pencil on a lined page.

Count what you can, date it, and name who saw it. If there was no measure, say so plainly, and ask which measure will be used next time.

A lot of real work never gets a number: calming a customer down, covering a gap nobody assigned, teaching someone the thing that isn't in the manual. It still has a count and a date. "Handled difficult customers" becomes "Took 31 escalated complaints from other associates between January and June and resolved 27 without a manager." If you didn't count at the time, start now, and say in the self-assessment that this year's figure is an estimate.

Then ask the person writing your review, in writing, what measure they'll use for that work next year. If the answer is none, you've learned how it's going to be weighed.

What do you do with work the criterion doesn't count?

Name it anyway, under its own heading, and ask how it'll be weighed. Work that's left out of the self-assessment is easy to leave out of the rating.

Malik co-leads the company's LGBTQ+ employee resource group. None of his goals mention it. "Led the ERG" tells the reader nothing. Written out, it says: "Outside my stated goals: co-led the LGBTQ+ employee resource group, about four hours a week. Ran the mentoring program that paired 22 newer employees with senior staff, and 18 finished it." Then the question, in the same document: "How will this work be weighed in my rating?"

A written criterion only covers what someone thought to write down. In interviews with 118 faculty of color at one research university, people described heavy service demands, often to make a committee look diverse. The authors describe that service as "not acknowledged or rewarded." Some had met every stated expectation and were still treated as if their work didn't count. One said of the new scoring rubrics, "everything they're coming up with I find more and more alienating" (Settles et al., 2019).

What if the rating criteria are vague?

A tape measure laid along a dark wooden surface, close up.

Ask for the definition, and for how it applies to your role, in writing. If it doesn't come, quote the vague word, write to the closest thing that's stated, and say what's missing.

A word like impact, leadership, or presence means whatever the person writing your review decides it means. What Readiness Costs follows one of those words through a career. Define the Goalpost is how to ask for the definition. Ask before you write the self-assessment, not after the rating.

Asking has a price. After Germany gave employees at workplaces with more than 200 people a legal right to ask how their pay was set, the most common reason women in one study gave for not asking was that "the employer might interpret the request negatively" (Göbel et al., 2026). That isn't a failure of nerve. The study has limits, though. It covers one law in one country. It analyzed gender as male or female only. And that answer, the most common, came from 23 percent of the women who hadn't asked. To better understand the true impact, any study would need to include all genders. Define the Goalpost sets out the price, and why the request still works where an argument doesn't.

Written criteria can change what gets said. At one U.S. technology company, managers began filling out a scorecard for each employee before the annual calibration meeting, with specific examples against each company value. Volunteer monitors and a timer kept the discussion consistent. The year before, 14 percent of women had been criticized in calibration as "too aggressive" and 8 percent of men as "too soft." At the first meeting after the change, those figures were 0 and 1 percent. Shelley Correll, who studied the change, also warns that formal procedures "can be misused by decision makers or, worse, have gender biases built into their design" (Correll, 2017). The criteria need checking too.

If the definition never comes, the gap goes on the record. Rae's line would read: "The criterion is 'demonstrates leadership.' It isn't defined for my role, so I've written to the two parts of my work that involve leading others: the returns training and the weekend schedule."

Should you rate yourself higher?

Not across the board. Rate yourself against the criterion, with the evidence beside each rating. A number with nothing beside it is easy to discount, at any height.

Women do rate themselves lower in some settings. In online experiments with more than 4,000 participants, women rated their own performance on a test about 13 points lower than men who'd scored the same, on a 0-to-100 scale. When participants were told their actual scores, the gap shrank but didn't close, and it was as large in private as when an employer would see it (Exley & Kessler, 2022). That was a test, not a workplace review.

The gap shows up at work too. In surveys at high-tech firms in Israel and the U.K., women and men reported the same manager ratings and gave themselves the same score when asked what it should have been. Asked to compare themselves with their peers, women rated themselves about five points lower on a 100-point scale (Baruch, 1996). The difference was in the comparison, not the work.

It isn't everywhere, though. In a meta-analysis of 99 samples, men rated their own leadership as more effective than women rated theirs, but most of that gap came from studies before 1996, and other people rated women as the more effective leaders (Paustian-Underdahl et al., 2014). In three studies of professional students taking an emotional intelligence test, people placed themselves at the 77th percentile on average when their actual average was the 41st, and gender made no difference (Sheldon et al., 2014). Plenty of people rate themselves too high.

And it isn't always about confidence. In two rounds of interviews with 59 Black women executives, the authors found women who already "exude high levels of confidence," and concluded that the usual advice to build it "may not apply" to them (Smith et al., 2019).

None of the studies in this section reports results for nonbinary people.

Writing to the criterion corrects in both directions. It gives an underrated result its evidence, and it holds an overrated one to the same standard.

What if writing it is the hard part?

If writing the document is itself the hard part, because of dyslexia, ADHD, depression, a brain injury, or a hand that can't type for long, you can ask to do it another way: more time, a conversation, a recorded answer. Whether to disclose is your decision. If you ask formally, this is what the law supports as of November 2026.

  • Which employers. The Americans with Disabilities Act covers employers with 15 or more employees on each working day in at least 20 weeks of this year or last year, including state and local governments. It doesn't cover tribal employers or tax-exempt private membership clubs (Americans with Disabilities Act of 1990, 2026). Federal agencies are held to the same standards under the Rehabilitation Act (Rehabilitation Act of 1973, 2026). State laws can reach smaller employers. New Hampshire's Law Against Discrimination covers employers with six or more employees, including the state and local governments. It doesn't cover a purely social club, or a fraternal or religious organization, if it isn't run for profit (Law Against Discrimination, 2026).
  • Who the law protects. Someone with a physical or mental impairment that substantially limits a major life activity, such as reading, concentrating, thinking, communicating, or working (Americans with Disabilities Act of 1990, 2026). A history of such an impairment counts too. Since 2008, the law has said that definition must be read in favor of broad coverage. A condition that comes and goes counts if it would substantially limit a major life activity when it's active. Whether it's limiting is judged without the help of medication, assistive technology, or ways of working around it you've learned (Americans with Disabilities Act of 1990, 2026).
  • How to ask. In plain language, out loud or in writing, at any point in your employment. You don't have to name the law or use the words "reasonable accommodation," and someone else, such as a family member or a health professional, can ask for you (U.S. Equal Employment Opportunity Commission, 2002). The federal appeals court for New Hampshire, Maine, Massachusetts, Rhode Island, and Puerto Rico has held that the request still has to be "sufficiently direct and specific" and explain how what you're asking for is linked to a disability (Reed v. LePage Bakeries, Inc., 2001).
  • What they can ask you for. If the disability or the need for the change isn't obvious, your employer can ask for reasonable documentation, limited to what shows you have a disability and need the change. In most situations, they can't ask for your complete medical records. They can't ask for documentation at all when both the disability and the need are obvious (U.S. Equal Employment Opportunity Commission, 2002).
  • Who sees it. Medical information your employer collects has to be kept confidential, and your employer can't tell coworkers you're getting an accommodation. The exceptions are narrow: supervisors can be told about the accommodation and any limits on your duties, first aid and safety staff can be told if you might need emergency treatment, and government officials checking compliance can be given what's relevant (U.S. Equal Employment Opportunity Commission, 2002).
  • What doesn't change. Your employer doesn't have to lower a production standard that applies to everyone, but may have to provide an accommodation that lets you meet it (U.S. Equal Employment Opportunity Commission, 2002). The standard you write to stays the same.
  • What they owe you. An accommodation that works, unless it would cause undue hardship, which means significant difficulty or expense, judged against that employer's resources and circumstances. Your employer can choose among options that work, but your preference should be given primary consideration (U.S. Equal Employment Opportunity Commission, 2002).
  • Asking is protected. Your employer can't retaliate against you for asking (Americans with Disabilities Act of 1990, 2026). The U.S. Equal Employment Opportunity Commission treats a request for accommodation as protected activity (U.S. Equal Employment Opportunity Commission, 2016), and the same appeals court has held: "Requesting an accommodation is protected conduct for purposes of the ADA's retaliation provision" (Freadman v. Metropolitan Property & Casualty Insurance Co., 2007).

The Job Accommodation Network's list of accommodations for executive functioning includes written instructions and performance standards stated clearly, such as a completion time or an accuracy rate (Job Accommodation Network, n.d.). A clear standard is something you can ask for, and Define the Goalpost covers how. If you're not sure which law covers your employer, ask your state's human rights agency or the U.S. Equal Employment Opportunity Commission.

Should companies stop asking for self-assessments?

There's a case for it and a case against. The research doesn't settle it, and either way the standard still has to be set.

The case for: self-ratings differ by gender and race, and the manager reads them. At one multinational financial services firm, women rated themselves lower than men on a 1-to-5 scale, and women of color rated themselves lowest. In 2016, a technical glitch meant managers set their ratings before they could see the self-ratings, and the managers' ratings lined up less closely with them.

The case against: in the same firm, rating without the self-ratings didn't move the gaps between groups. Managers appear to have leaned on the previous year's rating instead, and most of the race gap came from the managers' own ratings, not from the self-ratings (Bohnet et al., 2025). Hiding the self-ratings left the gaps where they were.

The authors stop short of telling firms to hide self-ratings. It could help people who don't have a rating history with the manager yet, they write, but where managers create the gap, the fix has to be aimed at managers.

Even the research that favors self-assessment puts conditions on it. A practitioner review argues it works when the criteria are clear, the person gets instruction and feedback, and there's a chance to revise (Kromrei, 2015). Baruch, whose surveys found self-ratings close to managers' ratings when employees already knew their manager's score, advised against using self-assessments to set pay (Baruch, 1996).

What does the person reading it owe?

The standard, stated at the start of the year and applied all year, and feedback in time to act on it. If those didn't happen, the self-assessment will show it.

For twenty-five years, I led front-line retail teams, and I read a lot of self-assessments. Most of them told me the same thing: no one had held the person to the performance standards consistently. They had nothing steady to write against, and that made developing them harder.

That's what a self-assessment shows the reader: whether anyone ever told the person what the work was supposed to be.

Feedback is part of the standard. One managing director in the study of Black women executives described managers who held back critical feedback because they feared the "angry black woman" stereotype: "if they think you're going to challenge them on anything, then they just won't give you the feedback. So you won't get it" (Smith et al., 2019). A person who never heard the concern can't write to it.

Credit is part of it too. In interviews with 65 women of color, a teacher described her contributions going unrecorded: "oftentimes what I say is not even on the minutes of the meeting" (Bhattacharyya & Berdahl, 2023). If your own notes leave someone out, their self-assessment may be the only place that work shows up. Read it that way.

Small gaps don't stay small. In a computer simulation of an eight-level company, a rating bias favoring men that explained 1 percent of the variation in scores left women in 35 percent of the top jobs (Martell et al., 1996).

So before you read a single self-assessment:

  • Write down the criteria you'll rate against, in plain words, and send them before anyone writes.
  • Give critical feedback to everyone, during the year, in time to act on it.
  • Read each self-assessment against those same criteria. Where you disagree, write your reasoning in the criterion's words.
  • Credit work that isn't in your notes. Ask where it came from before you discount it.
  • Check how the rating becomes pay.

That last one matters. At one large U.S. company, women, African American and Hispanic employees, and employees born outside the U.S. got smaller merit-based pay increases than U.S.-born white men with the same performance score, job, work unit, and manager. The raises ran 0.4 to 0.6 percent lower a year. The author calls that gap small but significant, and notes that over a career it can add up. After the company added accountability and transparency to its pay decisions, the gap narrowed and was no longer statistically significant. The study can't prove the change caused it, and the author says so (Castilla, 2015).

What a self-assessment can't do

It can't guarantee the rating. In the study of Black women executives, two of the three whose careers stalled were passed over after their success became visible (Smith et al., 2019). The record doesn't decide how it's read.

Write it anyway. A result written in the standard's words is on the record, and so is a standard nobody could find.


If you recognized yourself in this, or someone on your team, The Naming Work is where to start.

You already know what they call it when you do it. Nobody names the version being done to you.

Your work, your effort, and your discomfort get renamed one at a time, until the record reads like consent.

The Naming Work isn't a read. It's a scored diagnostic: fifteen questions, measured against a rights floor that you didn't set and that they can't move, not against how reasonable you're being.

You get a verdict. What you do with it is the actual work, through coaching or the app.

Name it before you change it.

Start at https://app.kovachconsultants.com/the-naming-work

References

Abraham, L. (2023). The gender gap in performance reviews. Journal of Economic Behavior & Organization, 214, 459–492. doi.org/10.1016/j.jebo.2023.07.039

Americans with Disabilities Act of 1990, 42 U.S.C. §§ 12102, 12111, 12112, 12203 (2026). law.cornell.edu

Baruch, Y. (1996). Self performance appraisal vs direct-manager appraisal: A case of congruence. Journal of Managerial Psychology, 11(6), 50–65. doi.org/10.1108/02683949610129758

Bhattacharyya, B., & Berdahl, J. L. (2023). Do you see me? An inductive examination of differences between women of color's experiences of and responses to invisibility at work. Journal of Applied Psychology, 108(7), 1073–1095. doi.org/10.1037/apl0001072

Bohnet, I., Hauser, O. P., & Kristal, A. S. (2025). Can gender and race dynamics in performance appraisals be disrupted? The case of social influence. Journal of Economic Behavior & Organization, 235, Article 107032. doi.org/10.1016/j.jebo.2025.107032

Castilla, E. J. (2015). Accounting for the gap: A firm study manipulating organizational accountability and transparency in pay decisions. Organization Science, 26(2), 311–333. doi.org/10.1287/orsc.2014.0950

Correll, S. J. (2017). SWS 2016 feminist lecture: Reducing gender biases in modern workplaces: A small wins approach to organizational change. Gender & Society, 31(6), 725–750. doi.org/10.1177/0891243217738518

Exley, C. L., & Kessler, J. B. (2022). The gender gap in self-promotion. The Quarterly Journal of Economics, 137(3), 1345–1381. doi.org/10.1093/qje/qjac003

Freadman v. Metropolitan Property & Casualty Insurance Co., 484 F.3d 91 (1st Cir. 2007).

Göbel, L., Weller, I., Nyberg, A. J., & Burkert, S. (2026). Gendered reactions to employer pay information nondisclosure. Academy of Management Discoveries, 12(1), 66–90. doi.org/10.5465/amd.2024.0013

Job Accommodation Network. (n.d.). Executive functioning deficits. Retrieved November 14, 2026, from askjan.org

Kromrei, H. (2015). Enhancing the annual performance appraisal process: Reducing biases and engaging employees through self-assessment. Performance Improvement Quarterly, 28(2), 53–64. doi.org/10.1002/piq.21192

Law Against Discrimination, N.H. Rev. Stat. Ann. §§ 354-A:2, 354-A:7 (2026). gc.nh.gov

Martell, R. F., Lane, D. M., & Emrich, C. (1996). Male-female differences: A computer simulation. American Psychologist, 51(2), 157–158. doi.org/10.1037/0003-066X.51.2.157

Paustian-Underdahl, S. C., Walker, L. S., & Woehr, D. J. (2014). Gender and perceptions of leadership effectiveness: A meta-analysis of contextual moderators. Journal of Applied Psychology, 99(6), 1129–1145. doi.org/10.1037/a0036751

Reed v. LePage Bakeries, Inc., 244 F.3d 254 (1st Cir. 2001).

Rehabilitation Act of 1973, 29 U.S.C. § 791 (2026). law.cornell.edu

Settles, I. H., Buchanan, N. T., & Dotson, K. (2019). Scrutinized but not recognized: (In)visibility and hypervisibility experiences of faculty of color. Journal of Vocational Behavior, 113, 62–74. doi.org/10.1016/j.jvb.2018.06.003

Sheldon, O. J., Dunning, D., & Ames, D. R. (2014). Emotionally unskilled, unaware, and uninterested in learning more: Reactions to feedback about deficits in emotional intelligence. Journal of Applied Psychology, 99(1), 125–137. doi.org/10.1037/a0034138

Smith, A. N., Watkins, M. B., Ladge, J. J., & Carlton, P. (2019). Making the invisible visible: Paradoxical effects of intersectional invisibility on the career experiences of executive Black women. Academy of Management Journal, 62(6), 1705–1734. doi.org/10.5465/amj.2017.1513

U.S. Equal Employment Opportunity Commission. (2002, October 17). Enforcement guidance on reasonable accommodation and undue hardship under the ADA. eeoc.gov

U.S. Equal Employment Opportunity Commission. (2016, August 25). Enforcement guidance on retaliation and related issues. eeoc.gov

This post is general information, not legal advice. The examples, including Rae, Dana, and Malik, are illustrations, not real people or data.

workplace systems · the practice

All posts