Event analytics skill

Event Analytics

Overview

Events are the atomic unit of customer understanding. Every signup, click, feature use, upgrade, and cancellation is an event. The mistake: tracking events haphazardly ("track everything!") with no taxonomy, creating a data swamp instead of a data pipeline. This skill covers event analytics across every GTM system — how to design an event taxonomy, implement tracking, and unify event data across your stack to build a complete customer picture.

Authoritative Foundations

• Segment — Customer Data Platform (CDP) and event taxonomy
• Amplitude — Behavioral analytics and event design
• Mixpanel — Product analytics and event-based reporting
• Intercom — Event-driven messaging and automation
• Avo — Event taxonomy and governance
• Snowplow — Open-source event pipeline

When to Use

Trigger phrases: "event analytics setup", "customer event tracking", "implement Segment", "Amplitude setup for SaaS", "event taxonomy design", "customer data pipeline", "product analytics events", "Intercom events", "event-driven workflows"

Step-by-Step Process

Phase 1: Event Taxonomy Design

The 4 event types (Segment spec):

1. Identify: Who is the user? identify(userId, traits)
2. Track: What did they do? track(eventName, properties)
3. Page: What page did they view? page(name, properties)
4. Group: What account/workspace are they in? group(groupId, traits)

Event naming convention (object-action framework):

Format: [Object] [Action] (Past tense. Write events as things that HAPPENED.)

Good: "Project Created" "Invoice Paid" "Campaign Sent"
Bad: "createProject" "user_clicked_button" "page3-conversion-v2"

Rules:
- Title case: "Signed Up" not "signed_up"
- Past tense: "Viewed" not "View"
- Object first: "Report Exported" not "Exported Report"
- No technical jargon: "Payment Completed" not "stripe_webhook_200"

Event taxonomy — core events every SaaS should track:

ACCOUNT EVENTS:
- Signed Up (properties: plan, source, referrer, UTM params)
- Subscription Started (properties: plan, price, billing period)
- Subscription Upgraded (properties: from_plan, to_plan, reason)
- Subscription Downgraded (properties: from_plan, to_plan, reason)
- Subscription Canceled (properties: plan, reason, tenure_days)
- Trial Started / Trial Converted / Trial Expired

USAGE EVENTS:
- Feature Used (properties: feature_name, context, duration)
- Search Performed (properties: query, results_count)
- Integration Connected (properties: integration_name)
- Invite Sent / Invite Accepted (properties: role)
- File Uploaded / Exported (properties: type, size)

ENGAGEMENT EVENTS:
- Email Opened / Clicked (properties: email_type, campaign_id)
- Notification Viewed / Clicked
- Support Ticket Created / Resolved (properties: category, priority)
- NPS Submitted (properties: score, comments)
- QBR Attended (properties: attendees)

REVENUE EVENTS:
- Invoice Created / Paid / Overdue
- Credit Card Added / Updated / Failed
- Refund Processed

MILESTONE EVENTS:
- Activation Complete (properties: time_to_activate_hours)
- First Value Achieved (properties: milestone, time_to_value_days)
- 7-Day Active / 30-Day Active
- Power User Threshold Reached

Phase 2: Implementation Strategy

Source → CDP → Destinations architecture:

Your App (client/server)
    │
    ▼
Segment (or Rudderstack / mParticle / Snowplow)
    │
    ├── Amplitude (product analytics)
    ├── Mixpanel (product analytics)
    ├── Intercom (event-triggered messaging)
    ├── Salesforce / HubSpot (CRM events)
    ├── Google Analytics (web analytics)
    ├── Data Warehouse (Snowflake / BigQuery / Redshift)
    └── Webhook → custom integrations

Implementation pattern (server-side preferred):

// Identify on login
analytics.identify(userId, {
  email: user.email,
  name: user.name,
  plan: user.plan,
  createdAt: user.createdAt,
  company: { id: workspace.id, name: workspace.name }
});

// Group for B2B (account-level context)
analytics.group(workspaceId, {
  name: workspace.name,
  plan: workspace.plan,
  employees: workspace.employeeCount,
  mrr: workspace.mrr
});

// Track key events
analytics.track('Feature Used', {
  feature: 'Email Finder',
  source: 'dashboard',
  credits_remaining: user.credits
});

Phase 3: System-Specific Event Analytics

Intercom Events:

• Use as event triggers for: onboarding tours (if user does X, show tour Y), email sequences (if user reaches milestone, send email), chat targeting (if user is stuck, offer help), in-app messages (if feature unused, promote it)
• Intercom('track', 'Feature Used', { feature: 'Reports' })

Salesforce Events (via Platform Events):

• Track: Lead Status Changed, Opportunity Stage Changed, Task Completed
• Use for: real-time dashboards, Slack alerts on key deals, enrichment triggers

HubSpot Events (via Custom Behavioral Events):

• Track: Form Submission, Meeting Booked, Email Clicked, Page Viewed
• Use for: lead scoring, workflow triggers, list membership

Amplitude / Mixpanel:

• Purpose: product analytics — not just what happened, but who did it, how often, and what happened next
• Key reports: retention curves, funnel analysis, behavioral cohorts, feature adoption, power user curve

Phase 4: Event-Driven GTM Automation

Example workflows:

Event: "Signed Up" → source = Google Ads
  → Add to Ad Conversions in Google Ads
  → Add to HubSpot as Lead
  → Enrich with Clearbit / LeadMagic
  → Route to SDR if ICP match
  → Send welcome email sequence

Event: "Feature Used" → feature = "Reports", 3rd time this week
  → Score: +10 engagement points
  → Trigger Intercom message: "Power user move: try Advanced Reports"

Event: "Trial Expired" → no conversion
  → Send re-engagement email sequence (3 emails over 7 days)
  → If no response: move to nurture list

Event: "NPS Submitted" → score < 6 (Detractor)
  → Create Zendesk ticket: "Follow up with Detractor"
  → Slack alert to CSM
  → Auto-schedule call with customer

Phase 5: Event Governance

Event dictionary (living document — reference per event):

Event Name	Properties	Trigger	Destinations	Owner
Signed Up	plan, source, referrer	POST /auth/signup	Segment → all	Eng
Feature Used	feature_name, context	Client-side track call	Amplitude, Intercom	Product

Governance rules:

• New events require: description, properties schema, destination list, owner
• No "track everything" — each event must have a purpose and an owner
• Deprecate, don't delete. Keep old events but stop sending them.
• Review event taxonomy quarterly. Remove unused events. Consolidate duplicates.
• Every event needs a test. CI/CD should verify events fire correctly.

Tools for event governance:

• Avo (avo.app) — event taxonomy design, code generation, validation
• Segment Protocols — enforce event schemas, block malformed events
• Amplitude Data — govern events within Amplitude's ecosystem

Output Format

EVENT ANALYTICS PLAN — [Company]

CDP: [Segment / Rudderstack / mParticle / Snowplow / Custom]

EVENT TAXONOMY:
[Docs link or table with event name, properties, trigger, destinations]

IMPLEMENTATION:
- Client-side: [SDK / library]
- Server-side: [SDK / webhook pipeline]
- Testing: [how events are validated]

DESTINATIONS:
| Tool | Purpose | Key Events |
|---|---|---|
| Amplitude | Product analytics | Feature Used, Signed Up |
| Intercom | Messaging automation | All usage + milestone events |
| HubSpot | CRM | Account events, revenue events |
| Data Warehouse | Analytics | All events |

EVENT-DRIVEN WORKFLOWS:
1. [Event] → [Action] — [trigger condition]
2. [Event] → [Action] — [trigger condition]

Implementation Checklist

• Event taxonomy documented with naming convention (Object-Action, past tense)
• 20+ core events implemented across Account, Usage, Engagement, Revenue
• Server-side tracking for critical events (not just client-side)
• Group/account-level context sent for B2B (workspace ID, plan, employees)
• Event dictionary maintained with owner, properties schema, destinations
• Event-driven workflows documented (event → action mapping)
• Event testing in CI/CD (events fire correctly, properties valid)
• No PII in event properties (email, name, IP — use pseudonymous IDs)

Quality Check

Before delivering, verify:

• Output matches the user's stated request
• Named frameworks or sources are reflected in the recommendation
• The deliverable is specific enough for an agent to execute
• Any assumptions, risks, or dependencies are explicit
• No unsupported claims, invented facts, or private/internal references are included

Common Pitfalls

1. Tracking everything. 500 events, 50 properties each, no one knows what any of them mean. The data swamp. Fix: Every event must have a purpose. Start with 20. Add as needed.

2. Inconsistent naming. signed_up, userSignup, Sign Up Completed all describe the same thing across different systems. Fix: One taxonomy. Object- action. Past tense. Documented in an event dictionary.

3. Client-side only tracking. Ad blockers block client-side tracking (30-50% of users). Critical events lost. Fix: Server-side tracking for key events (signup, payment, subscription changes). Client-side for behavioral events.

4. No group/account context for B2B. Events tracked to individual users but not linked to their company workspace. Can't answer "what are our top 10 accounts doing?" Fix: group() call on login linking user to workspace.

5. PII in event properties. email: "[email protected]" in event properties is a data privacy violation waiting to happen. Fix: Use user IDs. Store PII in your database, not your event pipeline.

Execution Artifacts

• references/framework-notes.md — named frameworks, citation anchors, and operating assumptions
• templates/output-template.md — copy-paste deliverable structure for the user
• scripts/check-output.py — local checklist validator for required sections This skill includes lightweight artifacts the agent can load on demand: Use the artifacts when the user asks for an implementation-ready deliverable, a repeatable workflow, or a quality check rather than generic advice.

Related Skills

• cs-analytics-dashboards — CS health scores and dashboards
• gtm-metrics — SaaS metrics stack
• campaign-analytics — Campaign performance analysis
• 1p-tagging-pixels — First-party tracking implementation
• a-b-testing — Experiment design and analysis
• attribution — Multi-touch attribution models