I've written the change note to `output/drivers-note.md`, with the script's figures in `output/drivers.json`. FY2026 EBITDA fell from $452K in Prior to $303K in Current, a change of ($149K), or (32.9%). The script's driver-by-driver walk adds up exactly to that change.

**What changed:** three of the six drivers moved.

| Driver | Prior | Current | Months changed | EBITDA effect |
| --- | --- | --- | --- | --- |
| Active customers | 46 | 44 | 12 | ($133K) |
| Gross margin | 26.2% | 26.6% | 12 | $39K |
| Headcount | 23.00 | 23.75 | 9 | ($54K) |

Revenue per customer, cost per head and other opex did not change.

**Why EBITDA moved:**
- **Fewer active customers is the main cause.** On its own it cut ($133K) from EBITDA, and it explains all of the ($509K), or (4.3%), drop in revenue.
- **Higher headcount added ($54K) of cost.**
- **A better gross margin offset part of the drop, adding $39K.**

The walk moves the drivers one at a time in a fixed order. These amounts are each driver's effect in that sequence, not its share of the total change.

**What the data can't answer:** the file shows how much each driver moved, not why. The note's Review section asks each owner which documented assumption supports their change:
- Which customers were removed.
- What raised the margin.
- Which hires were added, and when.

The assumption register would answer these if you'd like me to review it next.