What to Track

Overview

A strong analytics setup starts with the decisions you want to make, not the SDK calls. This guide helps you choose events, properties, and auto-event settings so you can answer questions about acquisition, activation, engagement, and retention for agentic and AI-native products.

The two pillars of data

Trodo builds user profiles from:

1. Events (what happened)

Actions in your product—captured automatically where enabled, and explicitly everywhere that matters:

Event type	Examples	Why track it
Navigation	Page views, deep links into a run	Content and entry-point performance
Product usage	Prompt sent, agent run started, tool approved	Feature adoption and workflow health
Conversions	Signup, subscription, export completed	Funnel progression
Engagement	Scroll, time on task, session depth	Quality of experience

2. User properties (who they are)

Stable traits and segments:

Property type	Examples	Why track it
Identity	`user_id`, email (via identify / people)	Return visits and support
Context	Plan, role, workspace, environment	Segmentation
Product state	Agent mode, model choice, integration count	Power users vs. evaluators
Lifecycle	Signup date, activation flags	Cohorts and retention

Use People and Identify to keep these fresh.

Planning your implementation

Step 1: Define your key questions

Acquisition

Which channels drive signups and activated accounts?
Which landing pages lead to first successful agent run?
How does paid vs. organic compare for qualified trials?

Activation

What share of new users complete the “aha” workflow in day one?
How long from signup to first value event?
Where do users stall in onboarding or templates?

Engagement

Which features correlate with weekly active use?
How often do users run agents vs. only browse?
What predicts expansion or seat growth?

Quality & trust

Where do users retry, edit prompts, or cancel runs?
Which error types are most common after a release?
How does satisfaction proxy (e.g. thumbs, export) trend by cohort?

Step 2: Map questions to events

Question: “Which campaign drives the most activated accounts?” Events

page_view with UTM (auto when auto-events are on)
signup_completed (custom)
agent_run_succeeded or your chosen activation event (custom)

Properties

utm_source, utm_medium, utm_campaign (auto when present on URLs)
plan, role (via people.set after identify)

Question: “Where do users drop between starting a run and a successful outcome?” Events

agent_run_started
tool_call_requested / tool_call_completed (as appropriate)
agent_run_succeeded / agent_run_failed

Properties

workflow_id, template_id, model, surface (web vs. embed)

Step 3: Choose auto-events

Auto-events are off until you set auto-events="true" (or enable them in code). Enable the bundle, then disable what you do not want:

Auto-event	Disable if…
`text_selection` / `copy_action`	Clipboard or selection is sensitive
`rage_click` / `dead_click`	You only want raw `element_click`
`page_performance`	You already capture Web Vitals elsewhere
`js_error` / `network_error`	Another pipeline owns client errors

<script>
  var script = document.createElement('script');
  script.src = 'https://cdn.trodo.ai/scripts/analytics/trodo.script.min.js';  
  script.setAttribute('site-id', 'your-site-id');
  script.setAttribute('auto-events', 'true');
  script.setAttribute('auto-events-disabled-events', 'text_selection,copy_action');
  document.head.appendChild(script);
</script>

See Auto-Events for the full catalog and runtime API.

Step 4: Plan custom events

// Agent workflow
Trodo.track('agent_run_started', {
  workflow_id: 'weekly_report',
  surface: 'in_app',
  model: 'gpt-4o'
});

Trodo.track('agent_run_completed', {
  workflow_id: 'weekly_report',
  duration_seconds: 42,
  outcome: 'success'
});

// Friction
Trodo.track('agent_run_failed', {
  workflow_id: 'weekly_report',
  error_class: 'timeout',
  retry_count: 1
});

Step 5: Identify and set traits

At login or account creation:

Trodo.identify(user.id);
Trodo.people.set({
  email: user.email,
  plan: user.plan,
  workspace_id: user.workspaceId
});

Naming conventions

Use consistent snake_case event and property names:

✅ Good
❌ Avoid

Trodo.track('agent_run_completed', { ... });
Trodo.track('integration_connected', { ... });
Trodo.track('onboarding_step_completed', { ... });

Trodo.track('AgentRunCompleted', { ... });
Trodo.track('agent-run-completed', { ... });
Trodo.track('run', { ... });  // too vague

Properties

Use snake_case: workflow_id, template_id, retry_count
Prefer stable IDs over display strings when breaking down in reports

Example event sets

Agent / copilot app

'prompt_submitted'
'agent_run_started'
'agent_run_completed'
'agent_run_failed'
'tool_approval_requested'
'tool_approval_granted'
'message_copied'
'export_generated'

Automation / builder SaaS

'workflow_created'
'workflow_published'
'run_scheduled'
'run_succeeded'
'run_failed'
'integration_enabled'

AI-enabled classic SaaS

'feature_used'            // with feature_name: 'ai_summarize'
'ai_suggestion_accepted'
'ai_suggestion_dismissed'
'subscription_upgraded'

Implementation checklist

Install the SDK

CDN, npm, or tag manager

Configure auto-events

Turn on and trim noise for your surfaces

Add custom events

Cover activation, success, and failure paths

Identify users

Call identify() when the account is known

Set people properties

Plan, role, workspace, and product flags

Verify in the dashboard

Live events and sample profiles

Next steps

Quickstart

Start sending data

Installation

CDN, npm, and GTM

INTRO

DATA IN

DATA MODEL

BEST PRACTICES

What to Track

Overview

The two pillars of data

1. Events (what happened)

2. User properties (who they are)

Planning your implementation

Step 1: Define your key questions

Step 2: Map questions to events

Step 3: Choose auto-events

Step 4: Plan custom events

Step 5: Identify and set traits

Naming conventions

Example event sets

Agent / copilot app

Automation / builder SaaS

AI-enabled classic SaaS

Implementation checklist

Next steps

Quickstart

Installation

INTRO

DATA IN

DATA MODEL

BEST PRACTICES

​Overview

​The two pillars of data

​1. Events (what happened)

​2. User properties (who they are)

​Planning your implementation

​Step 1: Define your key questions

​Step 2: Map questions to events

​Step 3: Choose auto-events

​Step 4: Plan custom events

​Step 5: Identify and set traits

​Naming conventions

​Example event sets

​Agent / copilot app

​Automation / builder SaaS

​AI-enabled classic SaaS

​Implementation checklist

​Next steps

Quickstart

Installation

Overview

The two pillars of data

1. Events (what happened)

2. User properties (who they are)

Planning your implementation

Step 1: Define your key questions

Step 2: Map questions to events

Step 3: Choose auto-events

Step 4: Plan custom events

Step 5: Identify and set traits

Naming conventions

Example event sets

Agent / copilot app

Automation / builder SaaS

AI-enabled classic SaaS

Implementation checklist

Next steps