Suede Analytics

Suede AI Analytics Skills

Read the worked delivery contract for a concrete input, expected artifact, failure response and validation example.

Compatibility

Works with event schemas, source and exports; browser/debugger and existing authorized analytics access for live firing proof. Configuration changes follow the production-change boundary in this skill.

Use this measurement playbook to build tracking that supports auditable marketing and product decisions.

Initial Assessment

Check for .agents/product-marketing.md (or .claude/product-marketing.md, or the legacy product-marketing-context.md) and read it if present: the key conversions, the decisions the data has to serve, and the tools already in place drive every recommendation here.

Then work the intake list under Task-Specific Questions below; ask only what the context file did not already answer.

Production Changes: Halt Before Mutating

Editing live tags, properties, destinations, or consent settings is the highest-consequence action in this skill, and Boundaries below forbids doing it without explicit authorization and a rollback plan. When a task requires one and you do not have both, halt in four parts:

1. Stop. Do not publish the container, edit the property, or change the consent configuration.
2. Name the blocker in one line ("publishing this GTM container version changes what fires for all live traffic; I have no rollback version identified").
3. Offer 2-4 options (stage it in Preview and hand over the trace; write the change as a diff for the owner to publish; publish after the user names the rollback version; scope the change to a test environment).
4. Wait for the answer. Do not pick one and continue.

The same halt applies to anything the Privacy and Compliance section below sends to legal or privacy review: an unresolved lawful-basis question blocks implementation, it does not get an assumption.

Tracking Plan Framework

Structure

Copy code
Event Name | Category | Properties | Trigger | Notes
---------- | -------- | ---------- | ------- | -----

Event Types

For comprehensive event lists: See references/event-library.md

Event Naming Conventions

Recommended Format: Object-Action

Copy code
signup_completed
button_clicked
form_submitted
article_read
checkout_payment_completed

Best Practices

• Lowercase with underscores
• Be specific: cta_hero_clicked vs. button_clicked
• Include context in properties, not event name
• Avoid spaces and special characters

Essential Events

Marketing Site

Product/App

For full event library by business type: See references/event-library.md

Event Properties

Standard Properties

Best Practices

• Avoid PII in properties
• Reuse the Standard Properties names above rather than inventing per-event variants

GA4 Implementation

Quick Setup

1. Create GA4 property and data stream
2. Install the Google tag (gtag.js) or GTM, with consent mode v2 defaults set before any tag fires
3. Enable enhanced measurement
4. Configure custom events
5. Mark key events in Admin (GA4 renamed "conversions" to key events in March 2024), then create Google Ads conversions from the key events that bidding should use

Event Example

javascriptCopy code
// GA4's recommended sign_up event, not a custom name, so it feeds built-in reports.
// The tracking plan can still call it signup_completed; map the name explicitly.
gtag('event', 'sign_up', {
  'method': 'email',
  'plan': 'free'
});

For detailed GA4 implementation: See references/ga4-implementation.md

Google Tag Manager

Container Structure

Data Layer Pattern

javascriptCopy code
dataLayer.push({
  'event': 'form_submitted',
  'form_name': 'contact',
  'form_location': 'footer'
});

For detailed GTM implementation: See references/gtm-implementation.md

UTM Parameter Strategy

Standard Parameters

Naming Conventions

• Lowercase everything
• Use underscores or hyphens consistently
• Be specific but concise: blog_footer_cta, not cta1
• Document all UTMs in a spreadsheet

Search and AI Assistant Traffic

Organic search and AI answer engines now send traffic that default channel groups misfile. Add a custom channel group in GA4 (Admin > Data display > Channel groups) with an AI assistants channel placed above Referral:

textCopy code
Session source matches regex:
chatgpt\.com|chat\.openai\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|claude\.ai|you\.com

Many AI-assisted visits arrive with no referrer and land in Direct, so pair the channel with a "How did you hear about us?" field (see suede-attribution) and with Search Console data for organic queries. Review the regex quarterly as assistants change domains. For the visibility work itself, route to suede-ai-seo and suede-seo-audit.

Debugging and Validation

Testing Tools

Validation Checklist

Each box closes on an artifact from the tools above, matched to the tool category's "Required current proof" in Tool Integrations below. An unchecked box does not mean "probably fine": it means the tracking is reported as unverified, never as done. Inspecting the tag config is not proof; a readback is.

• Events firing on correct triggers: a DebugView/live-events capture showing each event on the intended action
• Property values populating correctly: a property readback per event, values matched against the tracking plan
• No duplicate events: the same capture inspected for repeat fires (multiple containers, trigger firing twice)
• Works across browsers and mobile: the readback repeated on at least one non-primary browser and one mobile session
• Conversions recorded correctly: a source receipt plus a destination receipt for the conversion, not the source alone
• No PII leaking: the payload of a real captured event read field by field, plus masking/sampling settings for session replay

Report what was proven and what was not. "Instrumented" and "verified" are different claims; only the second one may cite this checklist.

Common Issues

Privacy and Compliance

Privacy, consent, retention, deletion, and identifier rules vary by jurisdiction, audience, data type, contract, and platform configuration. Do not treat this skill as legal advice or declare a universal consent rule.

Before implementation:

1. Identify the actual markets, audience age, data categories, vendors, purposes, and data flows in scope.
2. Review current official regulator and platform requirements for those jurisdictions and configurations; obtain qualified privacy or legal review when the requirement is unclear or material.
3. Document the approved lawful basis or consent state, retention and deletion behavior, access controls, and prohibited properties.
4. Collect only approved data, avoid direct personal identifiers unless the reviewed design expressly allows them, and test both allowed and denied consent paths.
5. For Google tags, set all four consent mode v2 signals (ad_storage, analytics_storage, ad_user_data, ad_personalization) before any tag fires. Google requires them for advertising features used with EEA traffic. Pattern and basic-versus-advanced trade-off: references/gtm-implementation.md.

Output Format

Tracking Plan Document

markdownCopy code
# [Site/Product] Tracking Plan

## Overview
- Tools: GA4, GTM
- Last updated: [Date]

## Events

| Event Name | Description | Properties | Trigger |
|------------|-------------|------------|---------|
| signup_completed | User completes signup | method, plan | Success page |

## Custom Dimensions

| Name | Scope | Parameter |
|------|-------|-----------|
| user_type | User | user_type |

## Conversions

| Key event | Event | Counting | Google Ads conversion |
|-----------|-------|----------|-----------------------|
| Signup | signup_completed | Once per session | Primary / secondary / none |

Task-Specific Questions

1. What tools are you using (GA4, Mixpanel, etc.)?
2. What key actions do you want to track?
3. What decisions will this data inform?
4. Who implements - dev team or marketing?
5. Are there privacy/consent requirements?
6. What's already tracked?

Tool Integrations

This pack does not ship analytics connectors. Use the user's authorized property UI, debugger, export, API, or installed connector and verify current official documentation before constructing a call.

Boundaries

• Do not claim an event, conversion, consent state, or attribution path works until a current debug or readback proves it.
• Do not mutate production tags, properties, destinations, or consent settings without explicit authorization and a rollback plan.
• Do not collect secrets, direct personal identifiers, or sensitive traits merely because a tool permits them.
• Do not decide business success from a single dashboard number; state the metric definition, window, denominator, and exclusions.

Routing

• Need experiment design or result interpretation -> use suede-ab-testing.
• Need paid-campaign decisions -> use suede-ads.
• Need attribution modeling, model comparison, or cross-tool reconciliation -> use suede-attribution.
• Need pipeline and CRM attribution -> use suede-revops.
• Need organic visibility diagnosis -> use suede-seo-audit.
• From those skills, route instrumentation plans and firing verification back to suede-analytics.