name: marketing-analytics description: Track marketing KPIs, interpret content performance over time, and recommend optimizations using temporal-summary and vault-linked reports. compatibility:

• Cursor
• Claude Code
• Any runtime that can run a CLI and read this skill allowed-tools: []

Marketing Analytics

When to use this skill

• User wants performance reviews, week-over-week deltas, or optimization recommendations tied to published work.
• User has (or will paste) metrics tables into notes; you never fabricate numbers—only interpret supplied data or vault-stated values.
• temporal-summary is the primary MCP prompt for time-bounded narrative over note changes and logged events.
• User wants experiment readouts (A/B subject lines, creative variants) tied to drafts/ variant headers.
• User is building a monthly deck from vault-native tables only.
• User wants funnels tied to URLs: knowtation search "https:// OR path /signup" --folder published --limit 20.

Role and responsibilities

• Connect metrics to creative and distribution decisions via paths (published/, drafts/, outlines/distribution-*.md).
• Produce actionable recommendations: hypothesis, expected impact, measurement, owner—each labeled data-backed vs hypothesis.
• Maintain a clean audit trail of what changed in the narrative when recommendations update strategy.
• If numbers are missing, MCP knowledge-gap must output a data request list before recommending budget shifts.
• Use memory-informed-search for “last time we reported this KPI, what definition did we use?”

Workflow

1. Published index: knowtation list-notes --folder published --order date --limit 50; knowtation get-note published/<entry>.md for URLs and campaign tags.
2. Time window narrative: MCP temporal-summary for the review period (e.g., last 14 days); MCP daily-brief for quick anomaly surfacing.
3. Hypothesis linking: MCP causal-chain only when user supplies plausible events (e.g., “email sent → spike”); otherwise frame as correlation only.
4. Content performance context: MCP search-and-synthesize: “Which messages appeared across notes during the surge?” with --folder drafts or published via CLI filters as needed (knowtation search "headline OR CTA" --folder drafts --limit 20).
5. Memory of past reviews: MCP memory-informed-search for “last month’s KPI definitions”; MCP memory-context for stakeholder preferences.
6. Gaps: MCP knowledge-gap → missing UTMs, missing baseline, missing channel split—feed back to distribution and writer.
7. Report note: knowtation write research/analytics-<YYYY-MM-DD>-<scope>.md or propose via POST /api/v1/proposals if canonical.
8. Refresh index: knowtation index after importing large metrics Markdown so temporal-summary sees new material.
9. Daily pulse: MCP daily-brief during active campaigns; MCP resume-session when analysis spans multiple working sessions.
10. Entities: MCP extract-entities on vendor export filenames or pasted dashboards to normalize campaign names across notes.

Output conventions

• Report: research/analytics-<date>-<campaign>.md sections: Snapshot, Top movers, Underperformers, Experiments, Recommendations, Data sources (paste locations, spreadsheet names, API export dates—no secrets).
• Frontmatter: tags: [analytics, report], project, period_start, period_end, related_published: (list of paths).
• KPI table: Markdown table; units explicit; note confidence (high/medium/low) per row.
• Experiment log: Subsection Experiments with ID, Hypothesis, Start/End, Result, Next step; link drafts/ variants where applicable.
• Caveats: Always include Data limitations (sample size, attribution model) when user-supplied metrics are partial.

Handoff patterns

• From marketing-distribution: Receive published/ URLs, campaign table paths, and scheduling changes (compare to temporal-summary).
• To marketing-strategy: Summarize what worked with evidence paths; attach knowledge-gap for broken tracking.
• To marketing-writer: Provide test hypotheses (headline/angle) mapped to measurable events; cite content-plan-eligible topics for rewrites.
• To marketing-editor: Flag copy elements correlated with low engagement as hypotheses, not verdicts.
• Hub: Ship canonical reports via POST /api/v1/proposals when finance or leadership reviews analytics notes.
• Broader synthesis: MCP project-summary when the user asks how analytics fits the whole GTM program—not just one campaign.
• Ingest: Paste CSV summaries into research/ working notes, then knowtation index before temporal-summary over that week.
• API: Publish signed-off reports through POST /api/v1/proposals; avoid POST /api/v1/notes for numbers others will cite.
• Consistency: knowtation search "KPI OR metric definition" --folder research --limit 15 before renaming columns in the KPI table.