Performance Reviews People Actually Finish: What We Stopped Writing and What We Brought Back

A standard performance review at a mid-size company looks roughly like this. The manager spends 3–4 hours on each report: gathering feedback from cross-functional peers, re-reading old 1:1s, formulating ratings across 12 competencies, aligning with HR, adjusting for "calibration." The document comes out to 8 pages. The employee opens it, flips through, looks at the final rating, reads the first two paragraphs — and closes it. The whole thing takes 6 minutes.

The 4-hours / 6-minutes ratio isn't "employee laziness." It's a signal that something is systemically broken in the format itself. If the person the document is aimed at doesn't finish it, the document isn't doing its job. And no amount of "let's make the employee more engaged" will fix that. What needs fixing is the format.

This article is about what we stopped writing in performance reviews on our team, what we brought back, and what one-page template actually works. It's a practical how-to, without the ideology of "let's abolish performance reviews."

Why the Typical Review Goes Unread

Before fixing — a diagnosis. A few reasons most performance reviews remain unread artifacts.

Too much content, too abstract. A typical review rates 10–15 "competencies" — leadership, teamwork, communication, proactivity, learning ability, and so on. Each with its own 1–5 scale and a comment. The employee looks at this and sees an evaluation broken into details with no overall picture. What matters most is unclear.

Looking backward, with no view forward. Most reviews are written in the format of "here's what you did in the first half-year." That's useful information, but it ages quickly: by the time it's read it's two months past the end of the period, and the employee is already thinking about the next sprint, not the last quarter. If the review has no forward view, it feels like dust.

Bureaucratic language. "The employee demonstrated a high level of competency in the area of stakeholder engagement." This phrase passed through 3 layers of HR sign-offs and lost a layer of meaning in each. In its place could have been "Anna negotiated with the clients herself on two crisis projects without escalating — that's a senior level." But that's harder to write.

The rating on the first page. Many reviews open with the final number — "3.5 out of 5," "exceeds expectations," "meets expectations." After that number, there's no point reading further: in a second the employee understands "what they got," and the context no longer matters.

No "what's next" section. A good review answers three questions: what went well, what didn't, what to do next. Most reviews answer only the first and sometimes the second. The third — the most valuable to the employee — gets skipped "because the cycle is closing."

All five reasons add up to one: the review is written for the system (HR, calibration, reporting), not for the employee. That's why the employee doesn't read it.

What We Stopped Writing in It

When we rebuilt the format on our team, we deliberately threw out several familiar parts. It's scary — because the HR industry has said for 30 years that they're needed. But without them it got better.

Competency ratings. Not 12 competencies × rating × comment. If a senior developer on my team is excellent at systems design but average at communicating with cross-functional peers, I don't need to rate them on a 5-point scale for "communication" across 12 cuts. I need to tell them "here's one specific moment where communication with peers let you down, let's work on that direction." More concrete, shorter, more useful.

The final numeric rating in a prominent place. If a rating is needed (for calibration, for pay), it lives in the HR system, not in the review document. There's no number in the review itself. This immediately changes the psychology of reading — the employee doesn't scroll to the number, they read it like a conversation.

Bureaucratic language. Any text a robot could write from a template, we throw out. If "demonstrated a high level of skill" can be replaced with "pulled two projects through in December, especially X," we replace it. The concrete is ten times more valuable than the abstract.

A "development areas" section with no concrete actions. "Development areas: communication" isn't a development area, it's a diagnosis. A development area is "next quarter, try in three situations to surface problems immediately rather than sitting on them for a week." With a concrete first step and who helps.

Sign-off by three layers. One review is signed by the manager and the employee. Not "manager → HR partner → calibration → another pass → another signature." Each additional approver lowers the quality of the text: adds bureaucratese and removes the human element.

This is very inconvenient for a large company with established processes. And it's very useful for an employee who actually reads the review.

The "Context — Behavior — Effect" Structure and a One-Page Template

What we brought back instead of what we threw out is a simple four-block structure. It fits on a single A4 page.

Block 1. Context — 4 lines at the top. The period (e.g. "January–June 2026"), the role ("senior developer"), the key projects the employee took part in. This isn't the evaluative part, it's setting the scene for the rest of the document.

Block 2. What went well — specifically. 2–3 situations with a real description. Not "worked great with the team," but "on project X in March she refactored authorization, which removed a bug that broke production once a week; rallied 3 colleagues around it, ran a review with the cross-functional team." Each situation — behavior + a measurable or observable result. ~8 lines.

Block 3. What didn't go well — specifically. 1–2 situations where the result fell short of expectations, and why. Not "needs to work on communication," but "on project Y in April she sat on design questions for two weeks instead of raising them in a meeting, and the release slipped by 5 days." The concrete makes feedback land rather than feel like an attack. ~6 lines.

Block 4. Next step — just one. Not "5 development directions," but one concrete action the employee takes in the next period, and who helps them. "On the next three projects, raise unclear design questions at the daily, don't sit on them. I (the manager) will keep this in mind and back you up in meetings. We'll see how it's going in a quarter." ~3 lines.

The whole page comes to about 25 lines, not counting headings. It reads in 3 minutes, not 6, and it gets read to the end, because there's no filler.

And the key point — the final rating isn't written in this document. If it exists in the HR system (a score, a bonus, a promotion), it's stored separately. The review itself is a conversation about work and development, not a payment order.

Where AI Helps and Where It Shouldn't

With AI assistants in performance reviews today, it's a delicate situation. Used wrong, AI becomes the very "mechanism that writes bureaucratic text for you," and quality drops sharply. Used right, it saves a manager several hours and improves quality.

Where AI helps:

• Gathering material. AI pulls together, for the period: the employee's achievements, mentions in recognition, progress on goals, activity in discussions. That's half an hour of work for the manager. Without AI — 1.5–2 hours.
• A draft by structure. Based on the gathered material, AI proposes a draft in the 4-block format. Not the final text — a starting point with concrete details that the manager edits.
• Wording hints. "This phrase is abstract — rephrase with a concrete example?" Not a dictate, a reminder.
• Tone checking. If a draft has too many "you should"s, AI can flag it and suggest a less directive phrasing.

Where AI shouldn't help:

• It doesn't generate the review for the manager. A draft — yes; the final text — no. If the manager didn't edit it, the review stays a draft and isn't sent.
• It doesn't evaluate the employee. A numeric rating, a ranking, a bonus recommendation — that's the manager's decision together with HR. AI doesn't assign ratings to people.
• It doesn't write "development areas" on its own. This is the most sensitive part; it requires knowing the context, the history of the relationship, an understanding of the person. AI can offer a sample phrasing, but choosing what to write must be the manager's job.
• It doesn't communicate with the employee. The final review comes from the manager's hand, not from AI. The employee must believe their manager wrote these words, or trust is destroyed in a single cycle.

The line runs like this: AI does the preparatory work, the manager writes the text and signs it. This rule is the same for all "human" documents — reviews, feedback, promotion announcements.

Quarterly or Annually — What Rhythm You Need

The last practical question — frequency. The old model is once a year. The new one is quarterly. Which to choose?

Once a year works if:

• The team is small and stable (< 30 people, low turnover).
• The business cycle is annual and changes little (consumer manufacturing, classic retail).
• Bonuses and salary reviews are tied to year-end and can't be done otherwise under corporate rules.

Quarterly works if:

• The team is growing or changes often (turnover above 15% a year).
• The business cycle is fast (startup, software product, sprints).
• You want real feedback, not "recalling something 11 months later."

In most cases a quarterly rhythm is better — it doesn't lag reality, and each review is shorter because the period is smaller.

The hybrid model we use ourselves: a quarterly review + an annual summary. Each quarter — a short review in the structure described above. Once a year — a short summary across the 4 quarters that no longer repeats the detailed material but draws conclusions and long-term directions. This gives the best of both worlds: timeliness and a strategic view.

What doesn't work is "no formal review at all." I've seen companies that dropped performance reviews "because they're outdated" and put nothing in their place. A year later they got: unclear grounds for bonus decisions, hurt employees who were forgotten and unrecognized, and managers who alone keep track in their heads of who contributes how much. Some structure is needed — just not the kind in the 2010s textbooks.

The Bottom Line

A performance review isn't "a formal document for HR," but a conversation about work and development in written form. If it doesn't get read to the end, it doesn't work — and the format is to blame, not the employee.

What we stopped writing in it: ratings across 12 competencies, the final rating in a prominent place, bureaucratic language, abstract "development areas" with no actions, sign-offs by three layers.

What we brought back: one page of 4 blocks (context / what went well / what didn't / next step), concrete situations with behavior and result, one chosen next step instead of five growth directions.

AI helps with gathering material and drafts; it doesn't write the final text and doesn't evaluate people.

A quarterly rhythm works better for most teams than an annual one.

And the key point — the review should get read to the end. If it doesn't, redo the format, don't try to "motivate the employee to read it."

In the next issue we'll go through AI prompts for HR that save a manager real hours — ten concrete templates you can start using this week.

If you'd like the one-page performance-review template, it's at the link in the card. No signup, copy it for yourself.