# Kenji's Ticket-Tier Divergence — Chapter 4 Worked Example

> Twelve months of tier-2 support ticket volume and mean resolution time behind Kenji’s Wednesday-mid-morning chart-family rebuild in Chapter 4 of *The Defensible Decision*. Nine months pre-automation, three months post-automation, one Big Idea about divergence.

**Wholly synthetic.** Any resemblance to a specific real support-operations dataset is coincidental; the numbers are constructed so pre-vs-post-automation averages land on the ~19%/34% split the Big Idea cites.

## Files

- [`kenji-ticket-tiers.csv`](kenji-ticket-tiers.csv) — flat rows: one row per month with volume, resolution hours, and pre/post window tag
- [`kenji-ticket-tiers.json`](kenji-ticket-tiers.json) — canonical scenario contract with source hash

## What it represents

Twelve months (Jul–Jun, fiscal-year order) of support tier-2 ticket volume and mean resolution time. March marks the tooling-automation go-live boundary. The pre-automation window (9 months, Jul–Mar) shows tier-2 volume averaging ~1290 tickets/month at ~3.3 hours mean resolution time; the post-automation window (3 months, Apr–Jun) shows tier-2 volume dropping to ~1050 tickets/month while resolution time climbs to ~4.4 hours. Chapter 4 uses the dataset to contrast Kenji’s Copilot-produced stacked-area chart (four tier series stacked on one canvas; divergence hidden inside the stack) with the rebuild as small-multiples (tier-2 volume in one panel, tier-2 resolution time in another, both on the same nine-month x-axis; divergence between the two lines IS the argument).

## Invariants (locked)

The scenario carries executable invariants that Chapter 4’s release contract enforces:

- Exactly 12 months in fiscal-year order (Jul, Aug, ..., Jun).
- Pre-automation window is 9 months (Jul–Mar); post-automation window is 3 months (Apr–Jun); automation month is March.
- Tier-2 volume drops ~19% pre-vs-post automation (measured pre-avg vs post-avg).
- Tier-2 mean resolution time rises ~34% pre-vs-post automation (same measurement).
- Big Idea text names both trends verbatim and cites both percentages.
- `stacked_area_failure` state declares 4 series and `argument_visible_in_five_seconds: false`.
- `small_multiples_rebuild` state declares 2 panels and `argument_visible_in_five_seconds: true`.

## Why the invariants matter

Chapter 4’s pedagogical point is that chart-family choice is downstream of the argument shape. If pre-vs-post volume drop landed at 5% or 40%, the «divergence» framing would soften or overreach. If the resolution-time climb were smaller than the volume drop, the «team eaten by harder tickets» frame would collapse. Locking the two percentages keeps the argument stable across the print figure, the practice-page solutions, and any downstream teaching material.

## What it does NOT represent

- A specific real support-operations tooling rollout.
- Tier-1, tier-3, or escalation volumes (the chapter contrasts one stacked-area chart with the rebuild; other tiers are noted in the scene but not modelled).
- The tooling vendor’s own metrics (the story is about Kenji’s analysis discipline, not the tool).

## Reproducibility

The dataset ships from the book source at `AlexBooks/applied-ai/ai-directed-viz/data/kenji-ticket-tiers.json` and is re-emitted here by `AlexBooks/applied-ai/ai-directed-viz/scripts/generate-kenji-tiers.mjs --site-root <site-root>`. The generator writes the source SHA-256 into `kenji-ticket-tiers.json` so drift between the book and this site is detectable.

## License

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