performance management
Performance Review Best Practices That Survive Real Manager Behaviour
Performance systems rarely fail on design. They fail on manager capability, weak evidence and a calibration step that never happens.
Hana Suzuki · 21 Jun 2026 · 8 min read
Ask any HR team what they changed last year and a redesigned review form is near the top of the list. Ask employees whether reviews improved and the answer is usually no. The form is almost never the constraint.
Get the cadence right
An annual review as the only feedback event guarantees recency bias and unpleasant surprises. The workable pattern is continuous lightweight one-to-ones, a mid-year checkpoint focused entirely on development, and an annual review that summarises evidence already discussed. Nothing in the annual review should be new information.
Evidence beats recollection
- Require managers to log specific examples throughout the year, not to reconstruct them in December.
- Collect peer input on defined questions rather than open-ended impressions.
- Tie assessments to agreed objectives and observable behaviours.
- Ask the employee for their own evidence first; it improves accuracy and reduces defensiveness.
Rating scales
Four or five points, with written behavioural descriptors for each level. Avoid forced distribution unless you can defend it; it damages trust in small teams and produces perverse calibration behaviour. If you use ratings at all, publish what each one means before the cycle starts.
Calibration is the step that creates fairness
- Group managers by level and function, and review proposed ratings together.
- Require each rating to be justified by evidence, out loud.
- Track distributions by gender, ethnicity, tenure and working pattern, and interrogate differences.
- Record the rationale for every rating that changes in calibration.
Separate pay from development
When the pay decision is in the room, the development conversation stops. Run the development discussion first, then communicate compensation in a separate conversation with a clear explanation of how it was determined.
Handle underperformance honestly
The most damaging pattern in performance management is the inflated rating given to avoid a difficult conversation, followed six months later by a performance improvement plan. If someone is not meeting expectations, say so early, in writing, with specific examples and specific support.
A review cycle where no manager changes their view during calibration is not a calibration; it is a scheduling exercise.
Frequently asked questions
- Should we abolish performance ratings?
- Only if you have a defensible alternative basis for pay, promotion and progression decisions. Organisations that remove ratings without replacing the decision framework typically make those decisions less transparently, not more fairly.
- How long should a performance review take?
- Budget sixty to ninety minutes for the conversation and a similar amount of manager preparation time. If preparation is under thirty minutes, the evidence is probably being reconstructed from memory.
- What is the most common performance review mistake?
- Recency bias, closely followed by rating inflation. Both are addressed by evidence logging through the year and a genuine calibration step.
Hana Suzuki
Talent Development Manager at Kaisei Industries
Calibration sessions, manager coaching and review cycles that people do not dread. I believe most performance systems fail on manager capability, not on the form.