Make Performance Reviews Work: Practical Moves That Improve Fairness and Actionable Feedback

The change that made our reviews fairer and lighter at the same time was making the evidence accumulate weekly instead of asking anyone to reconstruct a year from memory.

Most review processes are archaeology. Someone sits down in review week and tries to recall nine months of work, so what gets rated is whatever happened most recently and whoever was most visible. The usual fix is more form. More competency grids, more calibration meetings, and now the process is heavy and still unfair, because the underlying evidence never improved.

We are under ten people at Eprezto, so I have no calibration committee to hide behind. What we run instead is a weekly data-driven growth review. Whoever proposes an experiment owns it end to end and presents the result themselves, including the ones that did not work, mine included. By the time you want to assess someone there is a public record of what they took on, what they decided, and what it returned. The rating conversation stops being an argument about perception because everyone already watched the work.

Two things make that hold up. Peers, not only leadership, name the specific impact of someone's work in that review, which surfaces contributions a single manager would never see and forces recognition to be specific enough to check. And we reference only one dashboard, our single source of truth. If a number is not in it, it does not exist for the decision, which removes the round where two people arrive with different figures.

The honest part is that this works because ownership is real. If your people are executing someone else's plan, asking them to present results is theatre, and you are back to rating personality.

On the manager burden, the load did not move into a form. It moved to fifteen minutes a week, in public, where it is useful while the work is still live rather than filed six months after anyone can act on it.

A rating is either the last five minutes of something that ran all year, or it is fiction written in review week.

Six years of bootstrapping with small teams taught me that formal review cycles mostly benefit companies that can't have honest conversations the rest of the year. We scrapped the annual review entirely and moved to a single standing question asked every six weeks: "What's the one thing slowing you down that I could remove?" That shift alone cut the prep burden dramatically and made feedback land because it was tied to something the person actually cared about right now.

The one structural change that made ratings fairer was removing comparative scoring. We used to rate people on a 1-5 scale, which sounds objective but isn't. A 3 means something different depending on who's giving it and what mood the quarter was in. We replaced it with a simple three-bucket system: this person is growing into the role, performing well in it, or has hit a ceiling here. Those three buckets forced a real conversation instead of a number negotiation, and people stopped leaving reviews feeling like they'd just had their salary justified rather than their work assessed.

The failure that pushed us to change: one team member hit a 3 two reviews in a row with no clarity on what a 4 looked like. She left six months later. We realized the rating had communicated a vague dissatisfaction without ever giving her a path. The cost was losing someone who was actually good. After that, every bucket had to come with one specific next action, not a list of areas to improve. One thing.

Nobody trusts a review rubric their own manager wrote. So we didn't write ours. Our review cycle runs on three published, peer-reviewed instruments: the Individual Work Performance Questionnaire (is the work getting done, and does the person contribute beyond their own tasks), LMX-7 (how strong is each manager-report relationship), and the Team Climate Inventory (does the team itself function as a group). Three survey waves since February 2026, starting with our core operating team.

Nobody hand-computes a score, either. The survey database computes every score through 13 formula fields, trends grouped by month; the only thing written by hand is the narrative report. And each person's results are read against their own prior wave first, so the fight is with your own February number, not with your manager.

The change that mattered: a low score is a work order, not a verdict. A team lead's February observer ratings put three leadership components below our alert threshold, so we built interventions around exactly those three. By May every alert had cleared, his observer average had risen from 5.67 to 6.78, and team-average LMX-7 from 2.50 to 2.93. A small team and two readings — directional evidence, not proof, and I'd say so in print. But Gallup's 2023 workplace research finds only 8% of employees strongly agree their organization acts on survey results. We act on every wave. That's why people keep answering honestly.

Performance reviews became less of a burden the moment we stopped asking managers to write essays and started asking them to answer three fixed questions with one specific example each.

Our old review template had open text fields: strengths, areas for growth, overall comments. Managers dreaded it, reviews took two to three hours each to write, and the feedback that came out was often vague enough that the employee could not act on it, phrases like "needs to be more proactive" with no example attached.

We replaced it with three fixed prompts: one thing that worked this quarter with a specific example, one thing to change next quarter with a specific example, and one resource or support the manager will provide to help make that change happen. No prose beyond that.

Review time for managers dropped from roughly two and a half hours to under 45 minutes per person, and feedback quality went up, not down, because the format forced specificity instead of allowing vague adjectives to fill space. One account manager told me the new format was the first review in three years where she left the room knowing exactly what to do differently. That reaction, more than any efficiency metric, told me the change worked.

In a professional practice the honest starting point is that performance is already being reviewed continuously. Every job passes through review before it leaves the building, and the reviewer writes comments. So an annual review ought to be the summary of a record that already exists, not a fresh act of recall performed in one sitting.

That is where the burden and the unfairness come from at the same time. When a manager sits down to compose a review from memory, memory supplies the last six weeks and the two most emotional events of the year. The task feels heavy precisely because it is being invented rather than assembled.

The one change I made: nothing may appear in a review that was not said at the time it happened. If something is worth writing in the review, it was worth a sentence in the month it occurred, and if nobody said it then, it does not get introduced now.

That rule did three things. Ratings got fairer, because the year was sampled evenly instead of weighted toward whatever happened most recently. Feedback got actionable, because it arrived while the person could still act on it, which is the only moment feedback is ever actionable. And manager burden dropped, because writing became assembly. The expensive part of a review is composition, not typing.

Two supporting practices. Rate against a written description of what the level requires, stated in observable terms, rather than against other people. Most perceived unfairness is not favoritism. It is that one manager's meets expectations is another manager's exceeds, and employees compare across managers even when the company does not.

And hold the improvement conversation and the compensation conversation in separate meetings. Combined, only one of them is heard, and it is never the developmental one.

My fairness test is a single question: could this person have predicted their rating before walking into the room? If the rating is a surprise, the failure happened months earlier, and it belongs to the manager rather than the employee.

Running performance reviews that actually feel fair without bogging down management is a constant challenge, especially when you operate a highly automated, lean team like we do at Distribute. Traditional reviews usually mean managers spending hours filling out abstract 1-to-5 rating scales. It creates friction because team members naturally assume they are a top performer, and the ratings often feel like arbitrary choices made behind closed doors.

The single change we made that completely transformed our process was throwing out the standard performance rubric and replacing it with our balance sheet.

Instead of debating subjective behavior traits during a review, we open up a simplified version of our live cash flow. We show the team member exactly how their role fits into our current operational burn rate and the exact revenue milestones required to unlock the next tier of their compensation. The feedback portion then becomes entirely about what they specifically did to move us toward that next milestone, and what operational gaps are currently holding us back.

Anchoring the review directly to the company's live financial reality removes the burden of subjective grading for managers. It turns a tense, time-consuming evaluation into a shared look at the exact same puzzle, making the feedback instantly actionable.

We make all performance reviews fair and consistent for our administration staff as we have set the same goals and job description for all. Therefore, when a manager provides clear objectives at the beginning of the term, reviewing the progress becomes straightforward as the manager can review documented results.

The only thing that I did to make employee feedback much better is require an employee's self-review before a manager reviews. All administrative employees are required to do a five-minute survey on how well they completed their assigned tasks. The difference in what an employee says about their completion of tasks compared to what their manager thinks they accomplished indicates areas where there is a disconnect. As such, when you review your employee, you are no longer lecturing the employee but instead having a collaboration meeting where both parties agree on areas that need improvement. Feedback becomes transparent and easily translated into actions that improve the work of the administrative employee.

Performance appraisals remain limited to a few measures and concrete examples rather than being extensive documentation or complex rating process. The managers should be talking about their successes, difficulties and the help that would make the person perform better. Being practical in this situation is important as it will make the whole process more efficient and helpful for the employee without overburdening the managers with excessive administration.

What helped us improve the assessment and feedback was asking the managers to distinguish performance evaluation from impression and to justify each of their evaluations by citing concrete examples from the appraisal period. This made our calibration process more evidence based and minimized the effects of recency bias. In addition, the employees received much more concrete feedback since now they knew what exactly influenced their assessment and how to behave differently in the future.

Fairness and efficiency are balanced in administrative performance reviews when an emphasis is placed upon objective factual information as opposed to subjective judgments. The workload on managers has been reduced through a structured template used to evaluate employees' completion of their quarterly operational goals.

One of the changes we made that most improved the fairness of our evaluation process was weighting employee evaluations: 70% based on completion of objectives set forth by the employee's quarterly goals and 30% based upon the employee's qualitative demonstration of corporate values. Since the evaluation is so heavily weighted toward measurable data; this eliminates manager bias in evaluating employee performance. As a result, managers do not have to defend their evaluations and administrative personnel receive direct, fact-based feedback regarding their compliance with expectations and areas for future improvement.

We stopped mixing money talks with performance reviews at Wonderchat and it worked. Now we share ratings one week and discuss compensation the next. The difference is night and day. People listen to feedback instead of worrying about their raise. Managers can focus on coaching instead of defending numbers. It makes the whole process less tense and much more useful.

There are a few things that we do here, including keeping performance reviews to just twice per year and keeping them a bit more casual, but probably the most helpful change we've made over the years is how we now approach them from a more collaborative standpoint. Instead of just having our managers sit down with our employees and go through a list of hard checkpoints, telling them if they have or have not "succeeded," our managers mainly just have conversations with our employees. They'll talk about what they've observed, ask for insight from our employees, and work together to create goals for the next round of performance reviews. In getting feedback from both our managers and our employees, we know that this approach to performance reviews is preferred by everyone, because it allows for a lot more nuance and doesn't feel as stressful and high-stakes.

We reduce manager burnout in reviews through short forms and development discussions. Managers' need to provide lengthy, formulaic responses of page-long evaluations to all their employees is an example of how we delay meaningful feedback.

The one thing we did differently with regard to making the information we gave to employees about them, much more directly actionable, was enforce a 200-word limit and require managers to include a "single action plan" for each employee. In this action plan, managers are required to determine the single skill or workflow habit they feel will be most beneficial for the employee to acquire within the next six months. The limits we placed on length caused administrators to eliminate extraneous verbiage and provide actionable direction that could be implemented immediately.

To avoid manager fatigue in evaluations, and to provide fair evaluations for all back-office departments, we have created a common set of evaluation criteria based upon both objective completion metrics and a standardized competency rubric.

The one thing we did to make our calibration process more effective/faster is schedule short department-wide calibrations prior to finalizing ratings. Each department lead will meet and evaluate the preliminary ratings as a group to confirm that there are no inconsistencies in the way that ratings are being scored. These cross-departmental calibrations prevent each individual manager's tendency to be either overly lenient or overly strict when evaluating employees and creates a level playing field for all administrative personnel, creating complete confidence in the ability of managers to deliver objective, actionable feedback to their employees.

One change we made was separating performance discussions from final ratings. We found that people often focused on defending a score instead of understanding the feedback behind it. By creating a review stage centered on examples, outcomes, and growth areas first, conversations became more productive. Managers had better context before making evaluations and employees had a clearer view of what they could improve.

We also encouraged managers to document progress throughout the year instead of relying on recent events. This reduced bias and made reviews feel like a continuation of ongoing conversations rather than a stressful annual judgment. The process became easier to manage while creating more trust across teams.

We stopped mixing quarterly OKR reviews with our regular 1:1s at Joymore and it helped. Using a basic template across teams kept the conversations focused on goals and next steps. We tried a few other options, but this setup fit our remote team best. Just keep the format simple and check in often so it doesn't turn into a burden for managers.

Linking ratings to a few numbers and one learning goal worked better than anything else. It cut down the long write-ups and gave my teams something they could actually use. We tried a lot of different ways to do this, but focusing on real results was the only thing that stuck. When tech teams tracked NPS or speed, we could spot the problems immediately. Just stick to the numbers that matter and only speak up when something is clearly wrong. It keeps things fair and fast.

The change that did the most for both fairness and manager load was cutting the review to one question and forcing evidence to be gathered continuously rather than reconstructed at cycle time.

The one question: "what did this person ship in the last six months that only they could have shipped?" Every manager answers in about a page, with named artifacts and dates. That's the review. No 12-dimension competency rubric, no self-assessment questionnaire, no 360 survey with 40 respondents.

Two things improved immediately.

Fairness went up because the format is hostile to vibes. "She's really strong across the board" doesn't survive a question asking for specific work only that person could have done. Managers who couldn't answer discovered they'd been rating from impression rather than observation — exactly the failure mode that produces biased ratings, since impression tracks visibility and visibility tracks who talks in meetings.

Manager burden went down because the writing is short. The old system asked three hours per report. The new one takes forty minutes — but only if you've been keeping notes.

That's the second half: a running doc per direct report, one line added whenever something notable happens. Shipped something good, handled a hard conversation well, missed a commitment, unblocked another team. Thirty seconds, twice a week. At review time the evidence is already there, and the manager assembles rather than remembers.

The recency-bias fix was a side effect I didn't anticipate. Reviews built from memory over-weight the last six weeks badly. Reviews built from a running log describe the whole period, and people notice — several teammates said it was the first review that mentioned work from early in the cycle.

Actionability comes from a second short section: one thing to keep doing, one thing to change, each tied to a specific example. "Work on executive presence" is unactionable. "In the partner review last month, you had the right answer but opened with the caveat instead of the recommendation" is something a person can do differently next Tuesday.

Two things I'd avoid: forced distribution across teams (managers game it to protect people, calibration turns political), and mixing the rating conversation with compensation. Split those by two weeks and the rating gets honest.

One question, running notes, one keep and one change with examples.

The biggest burden on managers is subjective memory. If a review is based on how the day felt, two people doing the same job get scored differently and nobody can defend the number. The change that worked for me is tying ratings to the same checklist the job already runs on, not a separate form. Every clean my crews run follows a set task list, room by room, the same 113 items every time on a turnover job. When I review someone, I am not guessing how thorough they were. I am looking at what got checked off and the before and after photos that back it up. That turns the review into evidence, not opinion, and it cuts the time a manager spends writing it, because the data already exists from the work itself. The mistake I keep running into in service businesses is building the review on top of the job instead of inside it. Feedback gets more actionable too. Instead of telling someone to be more careful, I can point to the exact room or step that got missed. That is something a person can actually fix next time, instead of a vague number on a form.

I keep the formal review short by requiring evidence to be collected during the year rather than reconstructed from memory at the end.

Each rating must connect to an expected outcome, one observed example and the effect on the team or business. Descriptions such as "not proactive" or "excellent attitude" are not enough without behavior the employee can understand and change.

The improvement was adding a brief manager calibration before reviews were delivered. Only unsupported ratings and major differences between comparable cases required discussion, which reduced meeting time while improving consistency.

Every review ends with one capability to strengthen, one action and a date for follow-up. A long list of weaknesses rarely creates change.

The process feels fair when employees can see the evidence, respond to it and understand what stronger performance would look like in practice.

I made one change that managers resisted at first and later appreciated most: every rating now needs a future sentence attached to it. A score on its own can feel final and abstract. A sentence that begins with 'Next quarter, this would look stronger if...' turns the review into a coaching tool.

That made ratings fairer because inflated praise without direction stopped counting as good management. It also helped underperformers because criticism had to be specific enough to act on. The process became less burdensome over time since managers were no longer writing long narratives to justify vague numbers. One practical sentence created clarity, accountability, and a more useful path between performance and improvement.

I split our reviews into clinical safety and teaching impact tracks. The nurses and instructors finally get it. We rely on student feedback and incident logs for the scores now, so people see what is actually happening. In feedback sessions, just ask three questions: what to keep, improve, and try. It gives everyone a solid plan for the next cohort.