I've written the review to `output/register-note.md`, with the script's figures in `output/register.json`. The script's counts reconcile: flagged plus clean equals the total, both overall and by area.

**Headline:** Of 20 assumptions, 4 are flagged and 2 have a stale review.

**Findings**
- **A-04 Average headcount** is the most important gap. It changed from 23.00 to 23.75 in the Current version with no source recorded. It feeds the Headcount driver, so it changes opex. Elena Brandt owns it.
- **A-09 Inventory turns** (Tomas Herrera, last reviewed 2025-03-14) and **A-13 Line of credit interest rate** (Dana Whitfield, last reviewed 2025-06-02) have stale reviews.
- **A-16 Merit increase** has no owner. Someone needs to be named before the plan locks.

**Changes:** 5 values changed. Four have a source: active customers, gross margin, list price increase and days sales outstanding. Active customers and gross margin also feed planning drivers.

**Check these before you rely on the review:**
1. **The review date is 2026-01-01, not today.** That's the script's default. Today is 2026-10-03, so assumptions that were current in January may now be past the 180-day threshold. The real stale count is probably higher than 2. If you're locking the plan now, I can re-run it dated today, or on whatever lock date you want.
2. **Percentages are rounded in the report.** The CSV holds the line of credit rate as 0.0725, but the script shows it as 7.3%. Gross margin is similar: 0.2655 shows as 26.6%. Keep that in mind when you compare the review against the register or bank documents.