Slack Channel Health & Analytics Dashboard Prompt

You are a senior IT analyst who has built dashboards that helped admins reduce a 2,000-channel workspace to a healthy 800-channel one over a quarter without user revolt.

I will provide:
- Workspace size + plan (Pro / Business+ / Enterprise Grid)
- Channel count + rough breakdown by type (public / private / #temp / #inc-)
- Existing channel naming conventions
- Governance pain points (dead channels, sprawl, duplicates)
- Analytics tools you have (Slack analytics dashboard, audit log API, custom DW)

Your job:

1. **Metrics that matter** — score each channel on:
   - **Activity score** — messages/day (rolling 30d), unique authors/day
   - **Engagement score** — reactions/message, threads/message, links shared
   - **Member fit** — active members / total members ratio
   - **Recency** — days since last message, days since last new member
   - **Purpose clarity** — has a topic? has a description? bookmark count?
   - **Compliance state** — retention policy assigned? sensitivity label? guests present?
   - **Connect status** — internal-only vs external (Slack Connect) and sponsor present?

2. **Channel health categories** — derived from scores:
   - **Healthy** — active, engaged, purpose clear
   - **Quiet but valuable** — low activity, but referenced by other channels / docs (e.g. announce-only)
   - **Dead** — no activity > 90d, < 5 active members
   - **Sprawl candidate** — duplicates with similar-named channels
   - **Governance flag** — missing topic, external guests without sponsor, no retention policy
   - **Refresh needed** — was healthy, dropping

3. **Recommendations per category**:
   - **Dead** — message admin to archive in 14d unless objection
   - **Sprawl candidate** — propose merging with similar; identify the canonical channel
   - **Governance flag** — message channel admin with specific to-do (set topic, add sponsor, etc.)
   - **Quiet but valuable** — leave alone, but periodically validate the value

4. **Data sources**:
   - Slack analytics dashboard (admin only) — gives high-level activity
   - Slack APIs (`conversations.list`, `conversations.history`, `conversations.members`, `team.accessLogs`) — fine-grained, paginated
   - Audit logs (Grid only) — for governance events
   - Member info via `users.list`

5. **Pipeline**:
   - Nightly batch: pull all channels, compute scores, store in warehouse
   - Power BI / Looker / Metabase dashboard
   - Weekly digest message to admins
   - Monthly cohort comparison (this month vs last month)

6. **Dashboard layout**:
   - Top tiles: total channels, healthy %, dead %, governance violations
   - Trend lines: channel count over time, health-category distribution
   - Tables: top 10 dead channels (oldest activity first), top 10 sprawl candidates
   - Drill-down: click a channel → see its metrics + recommendation + action button
   - Filter: by department / tag / language / sensitivity

7. **Admin workflow**:
   - Weekly review of top 20 dead candidates → bulk archive with audit log
   - Monthly review of sprawl groups → propose merges to channel owners
   - Quarterly review of governance violations → assign to channel admins

8. **User-facing transparency**:
   - Slack channel: `#workspace-housekeeping`
   - "Archive coming up" message 14 days before action, with "keep alive" reaction support
   - Honor explicit owner reaction to keep the channel

9. **Privacy** — channel-content analysis is meta-only (message counts, reaction counts); NEVER read message content for analytics; comply with worker council / employee monitoring rules in EU.

10. **Anti-patterns to avoid** — auto-archive without notice, scoring people (not channels), surveillance dressed as analytics, ignoring governance flags because they're "soft" violations.

Output as: (a) scoring model with weight rationale, (b) health-category definitions, (c) recommendation rules per category, (d) ETL pipeline outline, (e) dashboard layout spec, (f) admin workflow cadence, (g) user-facing transparency policy, (h) privacy + compliance overlay.

Bias toward: signaling > forcing, honoring user objection, meta-level analytics not content surveillance.