HR & People

Performance Reviews & the Ombudsman Channel

Running performance review cycles in Fintra - self, manager, calibration, decision - and the Ombudsman: a confidential, anti-retaliation channel for employee concerns.

Updated 9 min read1 labHR AdminOwner / FounderEmployee

Two HR processes live or die on trust in the record: performance reviews (people accept outcomes they believe were fairly produced) and raising concerns (people speak up only where confidentiality is real). Fintra structures both - reviews as a staged cycle with calibration before decisions, and the Ombudsman as a confidential channel with access restrictions enforced by the system, not by promises.

The performance review cycle

Reviews run in cycles (e.g., "2026 Mid-Year"). Within a cycle, each employee’s review moves through fixed stages, and nothing becomes an outcome until calibration and decision.

The five stages

  1. 1

    Open the cycle

    HR creates the cycle with its period, participant set, and timeline. Reviews are created for each participant against the cycle.

  2. 2

    Self-review

    Employees write their own assessment first - accomplishments, misses, growth areas - before seeing any manager input. Writing first prevents anchoring.

  3. 3

    Manager review

    Managers complete their assessment with a proposed rating. AI assistance can draft structure from goals and recorded work, but the manager owns every sentence they submit.

  4. 4

    Calibration

    Ratings are compared across teams before anyone is told anything - the step that catches the tough-grader/easy-grader problem and rating drift. Calibration adjustments are recorded as such, distinct from the manager’s original.

  5. 5

    Decision and delivery

    Final ratings and any compensation outcomes are recorded as decisions with their maker, then delivered in a human conversation. The system holds the record; the manager holds the meeting.

The Ombudsman: confidential concerns

The Ombudsman is a standing channel where any employee can raise a concern - conduct, safety, ethics, interpersonal conflict - with confidentiality enforced by the system. Reports can be submitted anonymously; an AI triage step helps classify category and severity at intake; and read access to case details is restricted to HR, legal, and admin/owner roles.

  • Submission: an employee describes the concern; triage suggests category and severity and reminds the reporter of the confidentiality and anti-retaliation policy. The employee can mark the report anonymous on submit.
  • Access control: non-privileged users cannot list or read case details - the restriction is enforced at the API level, not by convention. Anonymous reports stay anonymous in every view.
  • Case handling: privileged handlers work the case - investigation notes, status, resolution - on a need-to-know basis. An AI case summary can assist the handler; it carries the same access restrictions as the case.
  • Oversight: a risk dashboard aggregates categories, severities, and statuses so leadership sees patterns (e.g., a cluster of scheduling-fairness complaints) without exposing individual reporters.

What employees can be told about confidentiality

The channel only works if employees believe it. This is the honest description to give them - accurate about both the protections and their limits.

QuestionHonest answer
Who can read my report?Only HR, legal, and admin/owner roles - enforced by the system’s access control, and handled need-to-know within that group
Can I report anonymously?Yes - mark the report anonymous at submission; it stays anonymous in case lists and dashboards
Can my manager see it?Not unless they are in the privileged handler group - line managers have no ombudsman access
Is there a log of who accessed my case?Case handling actions are recorded; access is auditable
Are there limits?Yes - legal obligations (e.g., safety threats, certain legal processes) can require disclosure, and anonymous reports may limit how fully something can be investigated
Confidentiality: what is and is not promised

Hands-on labs

Practice against a realistic scenario. Each lab lists the steps, what you should see, and the checkpoints that confirm you got the same result.

Lab 1

Run a calibration that changes a rating - correctly

Scenario

In Acme’s mid-year cycle, Luis rated technician Dana Okafor 3/5 while rating his whole team a full point below Jordan’s team average - the classic tough grader. Separately, an employee has submitted an ombudsman report about weekend-shift assignment fairness. Handle both with the record intact.

Steps

  1. 1

    Open the cycle dashboard after manager reviews close and compare rating distributions by manager.

    Expected: Luis’s team averages 2.8; Jordan’s averages 3.9 - a distribution gap worth calibrating, visible before anyone was told anything.

  2. 2

    Hold the calibration session and adjust Dana’s rating from 3 to 4, recording the rationale ("rating normalized across managers; Dana’s output matches 4-rated peers").

    Expected: The calibration adjustment is recorded separately from Luis’s original rating.

  3. 3

    Record final decisions for the cycle, then have managers deliver conversations.

    Expected: Dana is told 4/5 once, after calibration - never a number that later changes.

  4. 4

    Separately: as the ombudsman handler, open the shift-fairness report (submitted anonymously) and note the triaged category and severity.

    Expected: The case shows category/severity from triage; the reporter’s identity is absent everywhere.

  5. 5

    Verify access control: attempt to view the case list as a line manager (Luis).

    Expected: Denied - ombudsman cases are invisible to non-privileged roles.

  6. 6

    Work the case: investigate shift assignment data, record findings and the fix (rotation policy), and resolve the case. Check the risk dashboard.

    Expected: The dashboard reflects one resolved fairness case - as a pattern data point, not a person.

Checkpoints - you got it right if…

  • Dana’s original rating and the calibration adjustment both exist in the record, with rationale
  • No rating was communicated before calibration completed
  • The anonymous report stayed anonymous through triage, handling, and the dashboard
  • A non-privileged user was denied access to case data

Frequently asked questions

Who can read ombudsman reports, and is that enforced or just policy?

Read access to case details is restricted to HR, legal, and admin/owner roles, and the restriction is enforced by the system’s access control - non-privileged users cannot list or open cases. Within the privileged group, handling is need-to-know, and case actions are recorded and auditable.

Are anonymous reports really anonymous?

A report marked anonymous at submission carries no reporter identity in case lists, handler views, or dashboards. Two honest caveats belong in your policy: anonymity can limit how fully a claim can be investigated, and legal obligations (safety threats, legal process) can override confidentiality in rare cases.

Does AI decide anything in performance reviews?

No. AI can help draft manager review structure and summarize recorded work, and triage helps classify ombudsman intake - but ratings come from managers, adjustments from calibration, and outcomes from recorded human decisions. Every stage has a named human owner.

What stops a manager from retaliating against a reporter?

Three layers: managers cannot see ombudsman cases at all (so most never know a report exists), the policy states retaliation is prohibited and reports remind employees of it, and - the layer only you can provide - leadership reviews adverse actions against known reporters with heightened scrutiny. The system removes the information pathway; culture must do the rest.

Can calibration change a rating after the employee was told?

The cycle is sequenced so it cannot happen in order: decisions unlock after calibration, and delivery follows decisions. If something truly must change post-delivery (a fact error), record it as a new decision with rationale - the audit trail shows exactly what changed and why.

Ready to try it in your own workspace?

Fintra is the AI Finance Operating System for SMBs - accounting, payroll, planning, HR, and compliance under one login, with governed AI doing the heavy lifting.

Talk to us