# Two reports answering the same question, one of them telling you not to trust it

CodeHerder shipped two first-pass-rate reports. One told readers not to trust it. Removed for $70.52, across 45 files.

Source: https://codeherder.com/examples/retire-legacy-first-pass-rate/

Examples · Refactor

Deleted the legacy first-pass-rate report everywhere it lived in the product, and pointed the docs at the stage-signals report that already replaced it.

[Request access →](https://codeherder.com/get-started/#request-access)

Measured from this task's real merge.

CodeHerder had two reports for 'first-pass rate by model and stage': an old one, and the workflow-aware report that replaced it. The product's own docs and CLI already told readers to ignore the old one and use the new one instead.

What was asked for

> Identify the single most impactful source of unnecessary complexity in this repo, and ship one cohesive change that makes it meaningfully simpler.

Filed by a developer, routed by capability (go, react).

What the herd did

1. plan
2. code
3. review
4. merge
5. verify
6. done

What it changed

The merge touched 45 files (+195 −1077). Among them:

- `internal/evalsuite/handler.go`
- `internal/evalsuite/repo.go`
- `cmd/ch/quality.go`
- `app/src/pages/ReviewDebt.tsx`
- `docs/api.md`

What it cost

$70.52 in model spend, 625 agent turns, and 4h 13m from filed to merged. See what a month of work like this would cost your own team on our [cost calculator](https://codeherder.com/calculator).

A report that tells you not to trust it is surface area, not a safety net.