# GreenLeaf reporting brief v2 — carry the learning forward

**Status:** AI-assisted editorial draft; human publication/release approval pending. Fictional case,
not client-operating evidence. This is an additive update to [complete brief v1](brief-v1.md), not a
replacement with a shorter prompt. Read both together for the next cycle.

The business goals, KPI definitions, stakeholder responsibilities, attribution rules, target periods
and cadence in v1 remain unchanged. The new knowledge is separated by evidence class below.

## Checked data findings

- Paid-mobile completion is 3,780/8,400 = 45% in Week 1 and 2,352/8,400 = 28% in Week 2.
- The mobile skeptical/exploring starts mix changes from 50%/50% to 25%/75%.
- Using fixed Week 1 channel/audience/creative weights gives 35.33% for Week 2. Both composition and
  within-group movement contribute descriptively; all 12 mobile strata decline. No cause is proved.
- The first text-only AI response and its focused follow-up contain arithmetic errors. Reuse the
  checked metric pack and claim references, not unverified numbers from those responses.

## Reported fictional context — verification outstanding

The product owner reports a mobile quiz presentation revision beginning September 8 at 09:00 UTC.
The support lead reports difficulty finding the next control on a phone, without dates, counts or
linkage to paid cohorts. No deployment audit, rollout share or version-assignment evidence is supplied.
These reports prioritize an investigation; they are not verified release logs or a causal diagnosis.

## Next-cycle work order

| Owner role | Required evidence/action | Stop condition |
|---|---|---|
| Product owner | Reproduce the mobile flow and retrieve release/version/QA evidence | Missing assignment/logs prevents a version-specific causal claim |
| Media lead | Reconcile event meaning, cohort tagging and comparable device/audience/creative cuts | Inconsistent definitions or immature windows prevent a settled comparison |
| Content lead | Inspect message/post-click expectations after measurement/journey checks | No confirmed message problem means no claim of a winning or failed hook |
| Reviewer with current measurement knowledge and release authority | Check the decision and all stakeholder drafts before client-facing release | Approval absent means no external delivery |

## Hypotheses that must remain hypotheses

Presentation friction, tracking change, targeting/mix changes and other influences may contribute.
The support report does not establish prevalence or abandonment. No quiz-driven revenue loss,
incremental lift, profit, realized LTV, full CAC or monthly-business-target verdict has been established.

Revisit the context, rule and evidence requests at the next reporting cycle; remove stale assumptions
or keep them explicitly unresolved. Neither this update nor its reuse proves improved future output.