The performance cycle design playbook

Whether a review cycle is trusted is decided before it starts — in the design decisions about criteria, weighting, cadence and calibration. This playbook walks through those decisions in the order they should be made, with the trade-offs stated plainly. It assumes nothing about your current tooling: the design questions are the same in a platform or a spreadsheet; the platform just makes the answers enforceable.

Step 1: Decide what the cycle is for

A cycle that feeds compensation, one that drives development, and one that documents underperformance are three different instruments. Most organisations need all three outcomes but should not weight them equally in one event. Write down the primary purpose in one sentence and let every later decision defer to it. When stakeholders disagree about weighting or anonymity mid-design, the purpose sentence settles the argument.

Step 2: Choose criteria people recognise

Criteria fail in two directions: too generic ("communication, teamwork, initiative" — identical for a nurse and a network engineer) or too numerous (twelve dimensions scored badly instead of five scored honestly). Aim for four to six criteria per role family: two or three universal ones that carry your values, and two or three specific to the discipline. The test is whether an employee reading their criteria can name, unprompted, what doing well looks like in their actual job.

Step 3: Weight before you see any scores

Weighting is a values statement — how much of this role is delivery versus collaboration versus growth? Set it per role family when the cycle is designed, and freeze it. Weighting adjusted after scores exist is indistinguishable from rigging, and managers can tell. If goal outcomes are a criterion (they should usually be one, weighted 20–40%), decide that now too.

Step 4: Set a cadence you can sustain

The best cadence is the one that runs completely every time. A lightweight monthly or quarterly check-in — fifteen minutes, three questions, no scores — plus a scored half-yearly or annual review beats an elaborate quarterly scoring cycle that collapses by Q3. Check-ins keep the data flowing so the scored review draws on the whole period, not on recent memory. Be honest about manager capacity: every artefact you add to the cycle is multiplied by team size.

Step 5: Decide the reviewer mix per level

Manager-only review is fast and blind; full 360 is rich and heavy. Match the mix to the stakes: manager plus self-review as the floor for everyone, add peer input for collaborative roles, add upward feedback for people managers. Decide anonymity rules explicitly and publish them — who sees raw comments, who sees aggregates, what is attributed.

Step 6: Build calibration in from the start

Announce before scoring begins that scores are provisional until calibration. Then calibrate on distributions, not anecdotes: put each team's score distribution side by side, ask managers at the extremes to talk through their reasoning against the criteria, and adjust where the rating culture — not the performance — is the difference. Document what changed and why; the audit trail is what makes the final numbers defensible in compensation discussions.

Step 7: Communicate the design, not just the deadlines

Most cycle communication is logistics — dates and links. Trust comes from explaining the machinery: what the criteria are, how weighting works, who reviews whom, what calibration does to raw scores, and what the outcomes feed. Ten minutes explaining the design at kickoff buys more goodwill than any amount of chasing at the deadline. Employees do not need to love the process; they need to be unable to call it arbitrary.

Running it in MeritFlo

Everything above is configuration in MeritFlo rather than convention: criteria and weighting per role family, cycle cadence with automated reminders, reviewer mixes per level, calibration views with distribution data, and a complete record of every score and adjustment. Design the cycle once, and the system runs it the same way for everyone — which is, in the end, what fairness means operationally.