Analytics & Insights skill

Analytics & Insights

GA4 "AI Assistant" channel group (added 13 May 2026)

Google Analytics 4 added a new default channel group called "AI Assistant" on 13 May 2026 (GA4 channel groups doc). When a referrer matches a recognized AI Assistant (ChatGPT, Gemini, Claude, etc.), GA4 automatically:

• Categorizes the session under the AI Assistant channel group
• Sets the Medium dimension to ai-assistant

This is the attribution-side counterpart to the new GSC AI Performance Report (rolled out 3 June 2026 — see /digital-marketing-pro:gsc-ai-performance). Because the GSC AI report intentionally excludes click data, the GA4 AI Assistant channel is currently the cleanest path to attribute actual traffic coming from generative AI surfaces.

Recommended GA4 setup checks when onboarding a brand:

1. Confirm the channel group is live in the property. Newer GA4 properties get it automatically; older ones may need it to appear after Google's backfill completes. If the brand reports their channel reports look unchanged after 13 May, check explore reports filtered by sessionDefaultChannelGroup = "AI Assistant".

2. Add the AI Assistant channel to custom reports + dashboards — for any brand running an AEO program (/digital-marketing-pro:aeo-geo, /digital-marketing-pro:aeo-audit), the AI Assistant channel trend is now a primary KPI alongside organic search clicks.

3. Don't merge AI Assistant into "Organic Search" or "Direct". Some legacy reporting templates roll AI traffic into Direct (because referrers weren't always present) or Organic Search (because answer engines feel "search-like"). Both are misattributions now — the AI Assistant channel is the authoritative bucket.

4. Reconcile with aeo-audit outputs and the GSC AI report. Three data sources, three different views:

   • aeo-audit (synthetic probing) — what AI engines could say about the brand
   • GSC AI Performance Report — actual impressions in Google AI Overviews / AI Mode (no clicks)
   • GA4 AI Assistant channel — actual traffic from non-Google AI assistants (clicks materialized)

   A healthy AEO program shows growth across all three; divergence between them is a diagnostic signal.

The blind spot: Google's own AI surfaces are filed as Organic Search (checked 2026-10-04)

Google's default-channel definitions (support.google.com/analytics/answer/9756891) say three things:

• AI Assistant rule: "The medium exactly matches 'ai-assistant'". GA4 sets that medium (and the campaign (ai-assistant)) when the referrer matches its list of AI assistants. Google's channel description names ChatGPT, Gemini, Deepseek, Copilot and Grok; the 13 May 2026 release note also names Claude.
• The AI Assistant channel "excludes Google's AI Overviews and AI Mode".
• Organic Search is the channel for non-ad links in organic-search results, "including Google's AI Overviews and AI Mode".

So a click from an AI Overview or from AI Mode is counted as Organic Search. The channel page documents no dimension that separates it from a classic blue-link click. Consequences:

• Never report the AI Assistant channel as "all AI traffic." It is non-Google assistant traffic only.
• Never subtract or estimate an "AI Overviews share" of Organic Search from GA4 data. No first-party metric for it exists, and an estimate presented as data is a fabrication.
• When AI Overviews impressions rise in Search Console while organic clicks fall, report both numbers side by side. Say plainly that GA4 cannot attribute the clicks to AI Overviews or AI Mode.

One table per AI surface: what you can and cannot measure (checked 2026-10-04)

Surface	Visibility metric (first-party)	Click / traffic metric	Not available
Google AI Overviews + AI Mode	Search Console generative AI report: impressions by page, country, date, device (help)	Inside GA4 Organic Search, inseparable	Clicks, CTR, and queries in the AI report; an AI-only slice of GA4 organic
Copilot, Bing, and select partner AI experiences	Bing Webmaster Tools AI Performance: citations, grounding queries, Intents, Topics, Citation Share (your share of all citations shown for a grounding query), and Compare periods. Intents, Topics, Citation Share and Compare are in preview (Bing blog, 16 Jun 2026)	GA4 AI Assistant channel (Copilot is a named source)	Bing says Citation Share "does not expose competitor domains, represent traffic share, or assign quality scores to content"
Google Shopping in AI Mode / AI Overviews / Gemini app	Merchant Center AI performance insights: brand share of voice vs similar brands, across discovery, evaluation and purchase, plus popular product terms and specifications (Merchant Center help; blog.google, 20 May 2026)	Merchant Center / Ads conversion reporting as usual	Availability: Google said it is rolling out in the U.S., Canada, Australia, India and New Zealand "in the coming months" (as of May 2026). Confirm in the account before promising it
ChatGPT, Claude, Perplexity, Gemini app (answers)	No first-party citation report exists. Use synthetic probes (/digital-marketing-pro:aeo-audit, /digital-marketing-pro:geo-monitor) and label them as probes	GA4 AI Assistant channel	Any vendor-published citation or impression count

Report each row in its own units. Do not add Bing citations, Search Console impressions and GA4 sessions into one "AI visibility" number. They measure different events on different surfaces.

When to Use This Skill

Activate this module when the user's request involves any of the following:

• KPI Frameworks: Defining the right metrics and success measures for a business model, campaign, or channel
• Performance Reporting: Building weekly, monthly, quarterly, or campaign-specific reporting templates
• Anomaly Investigation: Diagnosing sudden drops or spikes in traffic, conversions, or other metrics
• Competitive Intelligence: Analyzing competitor strategies, share of voice, positioning, and performance
• Attribution Modeling: Determining how credit for conversions is assigned across marketing touchpoints
• Marketing Mix Modeling (MMM): Estimating the impact of each marketing channel on overall business outcomes
• Incrementality Testing: Designing experiments to measure the true causal impact of marketing activities
• Dark Social Measurement: Tracking and attributing traffic from private sharing channels (DMs, Slack, email forwards)
• Privacy-First Measurement: Adapting measurement strategies for a cookieless, privacy-regulated environment
• Dashboard Design: Structuring dashboards for different stakeholder audiences

Trigger phrases: "KPIs," "metrics," "reporting," "dashboard," "why did traffic drop," "anomaly," "competitor analysis," "competitive intelligence," "attribution," "marketing mix model," "MMM," "incrementality," "lift test," "dark social," "cookieless," "privacy-first," "ROAS," "ROI," "performance," "what happened to our numbers"

Brand Context (Auto-Applied)

Before producing any marketing output from this module:

1. Check session context — The active brand summary was output at session start. Use the brand name, industry, voice settings, channels, goals, compliance, and competitors shown there.
2. If you need the full profile, read: ~/.claude-marketing/brands/{slug}/profile.json
3. Apply brand voice — Formality, energy, humor, authority levels must shape all content tone and word choices
4. Check compliance — Auto-apply rules for brand's target_markets and industry using skills/context-engine/compliance-rules.md
5. Reference industry benchmarks — Consult skills/context-engine/industry-profiles.md for the brand's industry
6. Use platform specs — Reference skills/context-engine/platform-specs.md for character limits and format requirements
7. Check campaign history — Run python campaign-tracker.py --brand {slug} --action list-campaigns before planning new work
8. If no brand exists, say: "No brand profile found. Use /digital-marketing-pro:brand-setup to create one, or I can proceed with general best practices."
9. Check brand guidelines — If ~/.claude-marketing/brands/{slug}/guidelines/_manifest.json exists, load and enforce: restrictions.md for banned words, restricted claims, and mandatory disclaimers; channel-styles.md for channel-specific tone overrides (may differ from base voice); messaging.md for approved key messages, taglines, and positioning language; voice-and-tone.md for detailed voice rules beyond the 4 numeric scores. If producing content for a specific channel, channel style rules take precedence over base voice settings.

Do not ask the user for information that already exists in their brand profile.

Required Context

Before executing analytics work, gather:

1. Business Model: SaaS, e-commerce, lead gen, marketplace, etc. (determines the KPI framework)
2. Business Maturity: Startup, growth, scale-up, or enterprise (determines measurement sophistication)
3. Current Metrics: What is already being tracked? What tools are in use?
4. Analytics Stack: Google Analytics (GA4), ad platforms, CRM, BI tools, CDPs, tag managers
5. Data Availability: How much historical data exists? What granularity?
6. Reporting Audience: Who receives reports? (Exec/C-suite, marketing team, board, clients)
7. Known Issues: Any known data quality problems, tracking gaps, or recent changes?
8. Geographic Scope: Single market or multi-market (affects privacy regulations)
9. Privacy Constraints: GDPR, CCPA, ATT — what consent mechanisms are in place?
10. Specific Question: If investigating an anomaly, what exactly changed and when?

For anomaly investigation, prioritize speed. Ask for the specific metric, timeframe, and any known changes. For strategic measurement work, gather the full context.

Capabilities

• KPI Tree Generation per Business Model: Hierarchical metric frameworks that connect top-level business goals to actionable marketing metrics, customized for SaaS, e-commerce, lead gen, marketplace, subscription, media, and other models
• Standardized Reporting: Templates for weekly performance snapshots, monthly strategic reviews, quarterly business reviews, and campaign post-mortems — each designed for different stakeholder audiences
• Anomaly Detection and Root Cause Diagnosis: Structured diagnostic framework for investigating sudden metric changes — systematic elimination of causes (tracking issues, external events, algorithm changes, seasonality, competitive actions, internal changes)
• Competitive Intelligence Framework: Methodology for monitoring competitor activity across channels (SEO, paid, social, content, PR), estimating competitor spend, and benchmarking performance
• Marketing Mix Modeling (MMM) Guidance: Framework for understanding channel-level contribution to business outcomes, including data requirements, model design considerations, and result interpretation
• Incrementality Test Design: Experiment design for geo-based lift tests, holdout tests, conversion lift studies, and matched-market tests to measure true causal marketing impact
• Dark Social Tracking: Methods for measuring private sharing activity (link shorteners, UTM-equipped sharing buttons, dedicated landing pages, survey-based attribution) and estimating dark social contribution
• Cookieless Attribution: Privacy-first attribution approaches including server-side tracking, first-party data strategies, modeled conversions, media mix modeling, and probabilistic methods
• Privacy-First Measurement Stack: Complete measurement architecture designed for GDPR/CCPA compliance, iOS ATT, cookie deprecation, and evolving privacy regulations
• Dashboard Architecture: Stakeholder-appropriate dashboard design with metric hierarchy, visualization best practices, and alert configuration

Process

Primary Workflow: Measurement Framework & Reporting

1. Business Context & Goal Alignment

   • Classify the business model and maturity stage
   • Identify the north star metric (the single metric most tied to business value)
   • Map business goals to marketing objectives to tactical metrics (KPI tree)
   • Determine reporting audience and their decision-making needs

2. KPI Tree Construction

   • Start with the top-level business goal (revenue, growth, profitability)
   • Break into marketing contribution metrics (marketing-sourced revenue, CAC, LTV)
   • Decompose into channel-level metrics (channel CPA, ROAS, conversion rate)
   • Add leading indicators (traffic, engagement, pipeline, MQLs)
   • For each KPI, define:
     • Definition: Exactly how it is calculated (no ambiguity)
     • Source: Where the data comes from
     • Benchmark: Target or industry benchmark
     • Cadence: How often it is reviewed
     • Owner: Who is responsible for this metric
   • Limit the framework to 15-25 KPIs total — more causes metric fatigue and diluted focus

3. Reporting Template Design

   • Weekly Snapshot (for marketing team):
     • Key metrics vs. target (traffic, leads, conversions, spend, CPA)
     • Week-over-week trends with directional indicators
     • Top 3 wins and top 3 concerns
     • Action items for the coming week
   • Monthly Strategic Review (for marketing leadership):
     • Month-over-month and year-over-year performance
     • Channel contribution breakdown
     • Funnel conversion rate analysis
     • Budget utilization and efficiency metrics
     • Strategic insights and recommendations
   • Quarterly Business Review (for executive/board):
     • Marketing contribution to business goals
     • CAC, LTV, and payback period trends
     • Competitive positioning update
     • Next quarter strategic priorities
   • Campaign Report (per campaign):
     • Performance vs. pre-defined KPIs
     • Channel-by-channel analysis
     • Creative and audience performance
     • Learnings and recommendations

4. Anomaly Investigation Protocol When a user reports a sudden metric change, follow this diagnostic sequence:

   • Step 1: Verify the Data

     • Is the tracking code still firing correctly?
     • Did a tag manager change, consent tool update, or analytics filter change occur?
     • Check for platform outages or reporting delays
     • If data is corrupted, fix tracking first — do not analyze bad data

   • Step 2: Define the Anomaly Precisely

     • Which metric changed? By how much? Over what time period?
     • Is it all traffic or a specific segment (channel, device, geography, page)?
     • Did it happen suddenly or gradually?

   • Step 3: Check External Factors

     • Google algorithm update (check SEMrush Sensor, MozCast)
     • Industry news or seasonal patterns
     • Competitor activity changes
     • Platform policy or feature changes

   • Step 4: Check Internal Factors

     • Website changes (deployments, URL changes, redirects)
     • Content changes (published, removed, or modified)
     • Campaign changes (launched, paused, budget shifted)
     • Technical issues (site speed, server errors, mobile rendering)

   • Step 5: Isolate and Diagnose

     • Cross-reference the anomaly with the identified factors
     • Determine the most likely root cause
     • Estimate the impact and expected recovery timeline
     • Recommend corrective actions

5. Privacy-First Measurement Architecture

   • Audit current measurement for privacy compliance gaps
   • Design a measurement stack that works without third-party cookies:
     • Server-side tracking for owned touchpoints
     • First-party data enrichment strategy
     • Privacy-compliant consent management
     • Platform-native conversion APIs (Meta CAPI, Google Enhanced Conversions)
     • Modeled conversions for attribution gaps
     • Marketing mix modeling for channel-level effectiveness
     • Incrementality testing for causal validation
   • Create a transition plan from current state to privacy-first architecture
   • Account for iOS ATT impact on iOS-heavy audience segments

Reference Files

• kpi-frameworks.md — Business-model-specific KPI trees, metric definitions, benchmark databases, and north star metric selection guide
• reporting-templates.md — Weekly, monthly, quarterly, and campaign reporting templates with stakeholder-appropriate formatting and visualization guidance
• anomaly-diagnosis.md — Diagnostic decision tree, common root causes by metric type, verification checklists, and resolution playbooks
• competitive-intelligence.md — Competitor monitoring methodology, tool recommendations, benchmarking frameworks, and competitive response playbooks
• mmm-framework.md — Marketing mix modeling data requirements, model design guidance, result interpretation, and optimization recommendations, including Google Meridian 2.x (JAX default and experiment-calibrated priors in v2.0.0; declarative calibration/holdout specs in v2.1.0; checked 2026-10-04)
• incrementality-testing.md — Experiment design templates (geo lift, holdout, conversion lift), statistical power calculations, and result analysis frameworks
• dark-social-tracking.md — Dark social measurement methods, implementation guides for tracking private shares, and estimation models
• privacy-first-measurement.md — Cookieless attribution approaches, consent management architecture, server-side tracking implementation, and privacy regulation compliance guide
• clv-analysis.md — Customer lifetime value models (historical, cohort-based, predictive, contractual), calculation guidance, and application to segmentation and budget decisions
• dashboard-design.md — Three-tier dashboard architecture (executive, operational, campaign), metric selection per audience, and visualization best practices

Output Formats

Deliverable	Format	Description
KPI Framework	Document + spreadsheet	Hierarchical metric tree with definitions, benchmarks, owners, and cadence
Weekly Performance Report	Document / dashboard spec	Templated snapshot of key metrics, trends, wins, concerns, and actions
Monthly Strategic Report	Document / dashboard spec	In-depth analysis with channel breakdown, funnel analysis, and recommendations
Anomaly Diagnosis Report	Document	Root cause analysis with evidence, impact estimate, and corrective actions
Competitive Intelligence Brief	Document + spreadsheet	Competitor overview, channel analysis, share of voice, and strategic implications
MMM Readiness Assessment	Document	Data availability audit, model feasibility analysis, and implementation roadmap
Incrementality Test Plan	Document	Experiment design, sample size, timeline, hypothesis, and success criteria
Measurement Architecture	Document + diagram	Full measurement stack design with privacy compliance and implementation plan
Dashboard Specification	Document + wireframe	Dashboard layout, metric selection, visualization types, and alert rules

Edge Cases

Insufficient Data for MMM (<2 Years)

• Situation: User wants marketing mix modeling but has less than 2 years of consistent marketing data
• Approach: Be honest about the limitation — MMM requires sufficient time-series data to separate signal from noise. With less than 2 years: (1) Start collecting and structuring data now for future modeling. (2) Use simpler channel-level attribution as a bridge. (3) Run incrementality tests to get causal data on key channels. (4) Consider lighter-weight approaches like regression analysis on available data with clear caveats about confidence levels. (5) Build toward MMM readiness with a data collection roadmap. Do not attempt to build a full MMM on insufficient data — the results will be misleading and potentially harmful to budget decisions.

iOS ATT Destroying Attribution

• Situation: Significant portion of conversions are untrackable due to iOS App Tracking Transparency opt-outs, making attribution data unreliable
• Approach: Acknowledge the gap explicitly rather than pretending attribution data is still complete. Implement: (1) Platform conversion APIs (Meta CAPI, Google Enhanced Conversions) to recover some signal. (2) Server-side tracking for owned touchpoints. (3) Modeled conversions using platform statistical models (with appropriate skepticism about platform self-reporting). (4) First-party data matching where consent exists. (5) Marketing mix modeling as a complement to click-based attribution. (6) Incrementality testing for high-spend channels. (7) Survey-based attribution ("how did you hear about us?") as a qualitative check. The goal is triangulation — no single method is sufficient; combine multiple approaches.

Dark Social Dominating Referral Traffic

• Situation: Large portion of "direct" traffic is actually from private sharing (Slack, WhatsApp, email forwards, Discord) and attribution is blind
• Approach: Estimate dark social impact by analyzing "direct" traffic to non-homepage URLs (people rarely type deep URLs directly). Implement measurement improvements: (1) Add social sharing buttons with UTM parameters to track shared links. (2) Use link shorteners with tracking for shareable content. (3) Create dedicated landing pages for community/sharing use cases. (4) Add "how did you find this?" surveys to key conversion points. (5) Monitor content share velocity using social listening tools. (6) Accept that some dark social will remain unmeasured and build that uncertainty into reporting. (7) Consider investing more in dark-social-friendly channels (community, word-of-mouth, referral) even without perfect measurement.

Multi-Touch B2B Attribution Across 12+ Month Cycles

• Situation: B2B enterprise deals take 12-24 months with dozens of touchpoints across multiple stakeholders, making traditional attribution models meaningless
• Approach: Abandon pure last-touch or first-touch models — neither represents reality. Implement: (1) Account-based attribution that measures touchpoints at the account level, not individual level. (2) Influence-based reporting that shows which channels contributed to pipeline, even if they didn't "source" the deal. (3) Weight models toward time-decay with higher weights on recent high-intent touchpoints. (4) Use self-reported attribution from sales team and buyer surveys as a complement to digital tracking. (5) Measure channel effectiveness by pipeline velocity (does this channel accelerate deals?) not just by sourcing. (6) Accept that perfect attribution is impossible for complex B2B and focus on directional insights rather than false precision.

Regulated Data Handling

• Situation: User is in healthcare (HIPAA), financial services, education (FERPA), or other industries with strict data handling regulations
• Approach: Before any analytics implementation, flag the regulatory context. Ensure: (1) PII is never passed through analytics platforms without proper consent and processing agreements. (2) Data storage complies with regional requirements (data residency). (3) Consent management is explicit and granular. (4) Analytics vendors have appropriate compliance certifications (SOC 2, BAA for HIPAA, etc.). (5) User-level tracking is replaced with cohort or aggregate analysis where required. (6) Data retention policies are documented and enforced. Recommend involving a compliance officer or legal counsel for any measurement architecture in regulated industries. Never assume general analytics best practices are compliant in regulated contexts.

Related Skills

• Campaign Orchestrator — For translating analytics insights into campaign optimizations, budget reallocation, and strategic decisions
• Funnel Architect — For connecting funnel-stage metrics to the KPI framework and diagnosing conversion rate anomalies
• Content Engine — For measuring content performance, identifying content decay, and informing content strategy with data
• AEO/GEO Intelligence — For tracking AI visibility metrics and incorporating AI citation data into the measurement framework
• Audience Intelligence — For validating persona hypotheses with behavioral data and building data-driven segments
• Digital PR & Authority — For measuring earned media impact, backlink acquisition, and share of voice

Context efficiency

This skill's reference docs (skills/<this-skill>/*.md) sum to ~30-50KB. Don't load them eagerly — pick targeted sections:

• Grep before Read. Find the keyword or section heading first, then Read with offset + limit to pull just that range.
• Walk ${CLAUDE_SKILL_DIR} once. Use a single directory listing to see what's there, then Read only the files that match your current step.
• One source at a time. If the workflow says "consult three reference files," read them sequentially after deciding what you need from each. Bulk-loading all three blows the per-skill 5K-token budget that auto-compaction reserves.
• Strip noise from CSV inputs. If the input is a large CSV, grep the header line first to pick columns, then process row-by-row — do not Read the whole file into context.