Skip to main content

Create and manage review cycles

This guide is for owners, HR/People teams, and workspace leaders who run review programs.

If you get the setup right, everything downstream gets easier: clearer reviewer experience, fewer surprises during the run, and more defensible final decisions.

Design your cycle in this order​

1. Decide the decision​

Start with the decision this cycle should support.

Ask:

  • What decision will this cycle support?
  • Who needs to contribute to that decision?
  • What evidence should be comparable across people?
  • What context should stay narrative and human?

In practice:

  • Performance cycles usually need broader evidence and multiple reviewers.
  • Support or improvement-plan cycles usually need tighter prompts and shorter follow-up windows.
  • Lifecycle cycles usually need consistent timing and less manual coordination.
  • Promotion or readiness cycles usually benefit from panel input and stronger calibration.

2. Choose the right setup path​

In Performance > Reviews > Manage / Results, click Create Cycle.

  • Use a Template when you want a strong starting point quickly.
  • Use AI Assist when you know the intent but want help drafting stages and questions.
  • Build from Scratch when your process is already well defined.

All options open the same editor, so you can still reshape the cycle fully.

Templates in the product cover patterns such as:

  • annual, mid-year, quarterly, and monthly performance reviews
  • probation checkpoints
  • performance improvement plan reviews
  • success ratings
  • leadership 360-style feedback
  • promotion assessments

3. Configure the settings that change run behavior​

Open Review Settings and work through these intentionally.

Trigger and cadence​

You can run reviews in a few different ways:

  • Manual: send now, schedule for a date, or publish as a recurring cycle
  • Onboarding milestones: Day 1, 30, 60, 90, 120, 150, or 180
  • Offboarding milestones: Day 1, 30, 60, 90, 120, 150, or 180

Use manual when the timing is business-driven. Use onboarding or offboarding triggers when the timing should follow lifecycle events automatically.

Stage timing mode​

Choose one of two stage timing models:

  • One after another
  • All at once (blind)

Sequential timing works best when the order matters. Blind timing works best when you want independent judgment before people can see each other's submitted answers.

Audience​

You can target:

  • everyone in the workspace
  • selected departments
  • selected career levels
  • selected level sequences
  • specific people

If you use department, level, or level-sequence audiences, ClarityLoop builds the run from the people who currently match that audience.

For manual cycles with broad audiences, you can also exclude specific people.

In practice:

  • Career level means a specific assigned framework level, such as a named level inside a role or track.
  • Level sequence means the progression step number, such as everyone at Level 3 across different roles or tracks.

Use career level when you want a precise population. Use level sequence when you want a broader cross-role band.

Access​

You can add:

  • Additional owners who can operate the cycle
  • Leaders who can view results for their reporting line

Owners run the process. Leaders get visibility, not full operating control.

4. Architect stages before questions​

Stages drive review behavior more than question wording.

Each stage sets:

  • reviewer type (Self, Manager, Panel)
  • duration
  • stage instructions
  • for panel stages, how panel members are assigned

Good stage instructions tell reviewers what "useful evidence" looks like. They reduce cleanup work later.

5. Configure panel selection intentionally​

Panel stages support four assignment models:

  • fixed panel at setup
  • manager chooses later
  • reviewee chooses later
  • manager or reviewee chooses later

Use fixed panels when consistency matters most. Use choose-later models when the right reviewers depend on recent project work or collaboration context.

A few practical rules matter here:

  • panel selection must happen before the panel stage closes
  • panel members who already submitted cannot be removed
  • only the allowed chooser can update the panel for choose-later stages
  • if you use manager-chooses-later or reviewee-chooses-later, make sure the right person knows they own that step

6. Write for decision quality, not volume​

Question types available:

  • Performance Rating
  • Yes / No
  • Single Choice
  • Multiple Choice
  • Open Text
  • Info Block

Strong cycles usually mix structured scoring with concise written rationale. That gives you both comparability (for calibration) and context (for fairness).

You can also:

  • mark questions as required
  • allow optional free text on structured questions
  • control answer visibility for sensitive questions
  • use AI refinement to tighten wording before launch

Question visibility options are especially useful for sensitive prompts:

  • Standard
  • Reviewer only
  • Reviewer + reviewee's manager

That lets you keep most of the cycle transparent while protecting the few answers that need tighter sharing.

A simple formula for strong review design​

Most strong cycles use this mix:

  • a small number of structured questions for comparability
  • a small number of open-text questions for rationale
  • stage instructions that tell reviewers what evidence to include
  • durations short enough to create urgency, but long enough to finish well

If every question is open text, calibration gets messy. If everything is just scoring, nuance disappears.

Publish readiness and what can block launch​

A cycle is publish-ready when it has:

  • audience defined
  • at least one stage
  • stage questions present (or stage reuses previous-stage questions)

When you publish:

  • immediate manual cycles notify reviewers shortly after publishing
  • scheduled manual cycles notify on the selected date and time
  • recurring cycles continue on their configured cadence
  • onboarding and offboarding cycles send automatically at the selected milestone

Running a live cycle without losing control​

Use Run Overview from the cycle menu during active runs.

Run Overview is where you:

  • confirm the latest run timing
  • watch stage progress and timing
  • add reviewees after launch if someone was missed
  • jump into full results

Important behavior after first send​

For manual reviews, once the cycle has already gone out:

  • editing audience or start timing in settings affects a future send, not the already-running one
  • existing run responses stay preserved
  • if someone was missed in the current run, use Run Overview to add them instead of changing the cycle settings and hoping it backfills

Unpublish behavior​

Unpublishing stops future sends/notifications, but people already notified can still complete until their deadline window closes.

What becomes harder to change once a cycle is live​

ClarityLoop protects review integrity once a cycle is in use.

That means:

  • cycles with active use cannot always be reshaped freely
  • stages with responses can no longer have their reviewer model, duration, or fixed panel logic changed freely
  • questions with responses can usually only have text and answer visibility updated, not type or option structure

This is a good thing. It prevents accidental process drift in the middle of a live review.

Practical patterns by use case​

  • Performance cycle: self + manager + optional panel, with at least one common scoring question for calibration.
  • Probation cycle: short self + manager flow, focused on readiness, support needs, and checkpoint recommendation.
  • Improvement cycle: short recurring cadence, narrow prompts, and explicit evidence of change.
  • Promotion/readiness cycle: manager + panel with clear scoring anchors and tighter question visibility for sensitive prompts.

Common failure modes to avoid​

  • Overlong stages that reduce completion urgency.
  • Too many open-text prompts with no scoring anchors.
  • Panel “choose later” setup without clear ownership of panel selection.
  • Launching without clear stage instructions, then over-relying on reopen requests to fix quality.
  • Using blind mode when the real process actually needs staged handoff.
  • Publishing before audience, owners, and leaders are fully correct.
  • Treating finalization summaries as an afterthought instead of part of the decision record.

Next steps: