# Kenji's Ticket-Tier Divergence — Schema

Flat CSV in `kenji-ticket-tiers.csv`. One row per month.

## Columns

| Column | Type | Description |
| --- | --- | --- |
| `month` | text | Three-letter month code in fiscal-year order (Jul, Aug, ..., Jun). |
| `tier2_tickets` | integer | Count of tier-2 tickets resolved in the month. |
| `tier2_mean_resolution_hours` | decimal | Mean resolution time for tier-2 tickets in hours, to one decimal. |
| `window` | text | Either `pre_automation` (Jul–Mar) or `post_automation` (Apr–Jun). |

## Row inventory

- 12 rows — one per month.

Total: 12 rows.

## Deliberate design choices

The dataset carries both metrics on every row so the reader can reproduce either Kenji’s original stacked-area chart (which used tier-2 tickets alongside tier-1, tier-3, and escalations on one canvas) or the small-multiples rebuild (which uses tier-2 volume and resolution time on two panels sharing the nine-month x-axis).

The `window` column is redundant with the `month` column (given the automation-month convention) but explicit tagging makes the two windows testable in a single group-by without repeating the automation-month rule in downstream code.

Resolution time is stored to one decimal to match the graduated climb Chapter 4 walks through (each month adds ~0.1 hours until the automation window, where the increment jumps to ~0.2/month).

## License

Released for teaching use alongside *The Defensible Decision*. Use freely.