performance-check（绩效检查）

/dm:performance-check

Purpose

Pull live metrics from all connected analytics MCPs and produce a comprehensive performance snapshot. Compares current performance to KPI targets defined in the brand profile, previous-period benchmarks, and industry averages. Designed for quick health checks — run it daily, weekly, or on-demand to stay on top of marketing performance without switching between platforms.

Input Required

The user must provide (or will be prompted for):

• Time period: Today, this week, this month, this quarter, or a custom date range (e.g., "last 14 days", "Jan 1 - Jan 31")
• Channel focus (optional): Specific channels or platforms to prioritize (e.g., "paid search only", "email and social"). If omitted, all connected platforms are included
• Comparison period (optional): Period to compare against — previous period, same period last year, or custom range. Defaults to the equivalent previous period
• KPI targets (optional): Override targets for this check. If omitted, targets are pulled from profile.json goals and KPI settings
• Granularity (optional): Daily, weekly, or aggregate view. Defaults to aggregate for the selected period

Process

1. Load brand context: Read ~/.claude-marketing/brands/_active-brand.json for the active slug, then load ~/.claude-marketing/brands/{slug}/profile.json. Apply brand voice, compliance rules for target markets (skills/context-engine/compliance-rules.md), and industry context. Also check for guidelines at ~/.claude-marketing/brands/{slug}/guidelines/_manifest.json — if present, load restrictions. Check for agency SOPs at ~/.claude-marketing/sops/. If no brand exists, ask: "Set up a brand first (/dm:brand-setup)?" — or proceed with defaults.
2. Detect connected analytics MCPs: Check .mcp.json and active MCP connections to identify which platforms are available (google-analytics, google-ads, meta-marketing, linkedin-marketing, tiktok-ads, mailchimp, stripe, mixpanel, amplitude, shopify, etc.). Log any expected platforms that are not connected so the user knows about gaps in coverage.
3. Pull metrics from each connected platform: Request key metrics for the specified time period:

- Traffic: sessions, users, pageviews, new vs returning - Ads: impressions, clicks, spend, CPC, CPM - Conversions: leads, purchases, sign-ups, goal completions - Revenue: total revenue, average order value, transaction count - Engagement: open rate, click rate, bounce rate, time on site - Platform-specific: email deliverability, social reach, video views, app installs

1. Aggregate into unified dashboard: Normalize metrics across platforms into a single cross-channel view with consistent naming, currency conversion if multi-currency, and de-duplicated conversion counts where platforms overlap
2. Calculate KPIs vs targets: Compare actuals to targets from profile.json goals — flag green (on track or exceeding), yellow (within 10% of target), or red (missing by >10%). Include absolute and percentage variance for each KPI.
3. Compare to previous period: Calculate period-over-period change for every metric and attach trend direction (up/down/flat) with percentage change. If year-over-year data is available, include as a secondary reference point.
4. Benchmark against industry: Reference skills/context-engine/industry-profiles.md for the brand's industry to contextualize performance relative to category averages. Flag metrics significantly above or below industry norms.
5. Identify notable findings: Surface the top 3 wins (best-performing metrics or biggest improvements), top 3 concerns (underperforming or declining metrics), and any statistically significant changes that warrant deeper investigation.
6. Generate recommended actions: Based on the data, produce 3-5 specific, actionable next steps — e.g., "Pause underperforming ad set X", "Increase budget on high-ROAS channel Y", "Investigate traffic drop on Z", "Scale winning creative variant", "Run /dm:anomaly-scan for deeper diagnosis".
7. Save performance snapshot: Execute scripts/performance-monitor.py --brand {slug} --action save-snapshot to persist the snapshot for historical comparison and trend tracking across future runs.
8. Log significant insights: For any metric with a notable deviation, save via scripts/campaign-tracker.py --brand {slug} --action add-insight so findings surface in future reports and campaign planning.

Output

A structured performance snapshot containing:

• Executive summary: 2-3 sentence overview of overall marketing health with the single most important finding highlighted
• Channel-by-channel metrics table: Traffic, impressions, clicks, conversions, revenue, spend, CPA, ROAS, and engagement rate per platform — sortable by any column
• KPI scoreboard: Each tracked KPI with actual value, target value, percentage to target, variance (absolute and %), trend arrow (vs previous period), and RAG status (red/amber/green)
• Cross-channel summary: Total spend, total conversions, blended CPA, blended ROAS, total revenue, marketing efficiency ratio, and overall health assessment
• Period-over-period comparison: Percentage change for all key metrics vs the comparison period with directional indicators and sparkline-style trend data
• Industry benchmark context: How key metrics compare to industry averages from industry-profiles.md, with percentile ranking where data is available
• Notable findings: Top 3 wins, top 3 concerns, and any anomalies worth investigating further — each with supporting data points and severity indicator
• Recommended actions: 3-5 specific next steps with priority ranking, expected impact, and the platform or campaign each action applies to
• Data gaps: Any platforms that were expected but not connected, metrics that could not be retrieved, or time periods with incomplete data — so the user knows what is missing from the picture

Agents Used

• analytics-analyst — Metrics interpretation, KPI analysis, cross-channel normalization, trend identification, industry benchmarking, insight generation, and action recommendation
• performance-monitor-agent — Data aggregation from connected MCPs, baseline comparison, snapshot persistence, historical trend analysis, and gap detection