Calibration Session Prep Tool

The problem this kills

Calibration season runs on a fragile spreadsheet. Managers paste in proposed ratings, someone color-codes a distribution by hand, and the meeting itself turns into a blur of "let's move her from a 3 to a 4" with no record of why. A week later nobody can reconstruct who changed what, the 5s are quietly clustered in one team, and a high rating sits next to a write-up full of "needs improvement" - and nobody caught it in the room.

Worse, the final numbers get committed back to the HR system before anyone formally signs off. There's no clean before/after trail, no bias check, and if an employee or a regulator asks "how was this rating decided?", the honest answer is "we lost the spreadsheet version that mattered."

What you'll build

A small, login-protected web app that turns calibration from a spreadsheet scramble into a consistent, auditable session:

• Import proposed ratings for a whole team or department - from your review tool's CSV or a Google Sheet.
• See the picture instantly: a rating distribution chart, your target-distribution guidance overlaid (guidance, not a forced quota - unless you choose to make it one), and automatic outlier flags.
• Catch the red flags: high rating next to negative comments, ratings that drift far from level/role peers, and clustering patterns by group that hint at bias.
• Run the meeting inside the tool: every change from proposed to adjusted is captured with a required reason and the name of who made it.
• Lock it with a human gate: nothing becomes final until HR formally signs off the session. Then - and only then - the final set is committed and the full change history is frozen.
• Export a clean before/after CSV in the exact columns your system of record expects.

What's inside the Implementation Plan

The plan is a complete, paste-and-go runbook for an AI coding agent (Claude Code). You don't write code - you paste, answer questions, and approve.

It opens by interviewing you about your business - your review cycle, your rating scale, how your employee and level data is named, your typical and peak headcounts, your exact approval rules, and your messy edge cases - so the tool is tailored to how you actually run calibration, not a generic template. The agent reads a short spec back to you and waits for your thumbs-up before building anything.

From there it walks you step by step: the data model shaped to your rating scale, the import, the distribution and outlier views, the in-session adjustment capture, the HR sign-off gate, the audit trail, and the CSV export - each step ending in a ready-to-paste prompt.

The governance it includes (this is the point)

This isn't a toy. The plan bakes in the controls HR and audit actually care about:

• Login so only your calibration team can get in.
• Row-level security so each organization only ever sees its own employees and ratings.
• A complete audit trail - who changed which rating, from what to what, why, and when.
• A hard human-in-the-loop gate - the tool surfaces and drafts; a facilitator records adjustments and HR signs off; only then are final ratings committed.
• Duplicate guards keyed on employee ID + cycle, so the same person can't be imported or committed twice.

Who it's for

HR business partners and people managers who run calibration sessions today off a spreadsheet that everyone is a little afraid of. If you've ever finished a calibration meeting and couldn't fully explain how the numbers landed where they did, this is for you.

You've got this - paste the first prompt and let the agent interview you.