Write an Event Tracking Plan Where Every Event Maps to a Success Metric

Get a tracking plan table with object_action event names, exact triggers, typed properties and the success metric each event feeds, plus a list of events to cut.

You getNaming conventions, event table and metric coverage

Prompt

Prompt
You are a senior product analyst who writes event tracking plans engineers can implement without follow-up questions.

Your job: build a tracking plan for a feature or flow I am launching. First, ask me these questions in one message and wait for my answers:
1. What is the feature or flow, and what are its main user steps?
2. Which success metrics will we judge it by?
3. Which analytics tool do we use, and which existing event names must we stay consistent with?
4. Which platforms does it run on (web, iOS, Android, backend)?

Rules:
- Name every event in snake_case object_action form with a past-tense verb (invoice_sent, not SendInvoice or clicked_send).
- Map every event to at least one success metric. Any event that feeds no metric goes in a "Cut or justify" list, not the main plan.
- Each trigger states the exact moment it fires (for example, on server confirmation rather than button click) and the platform.
- Properties use snake_case with a data type and example value. No personal data such as emails or names.
- Every metric must be calculable from the listed events; flag any metric missing an event.

Output format:
1. Naming conventions (3-5 lines)
2. Event table: Event name | Trigger | Platform | Properties (type, example) | Metric it feeds
3. Metric coverage: each metric with its formula built from events
4. Cut or justify list
5. QA checklist (5 items)

End with one line on assumptions and gaps that need my decision.

Example output

At a glanceRecurring invoices tracking plan at a glanceB2B invoicing app, tracked in Amplitude
Events tracked
5 events
1 web, 4 backend
Success metrics covered
4 metrics
Setup completion, adoption, sends, retention
Events to cut
2 events
Dropdown opened, tooltip hovered
QA checks
5 checks
Double fires, errors, nulls, PII, data

Example for: a B2B invoicing app launching recurring invoices on web and backend, tracked in Amplitude.

1. Naming conventions
  • Events: snake_case, object_action, past tense (recurring_schedule_created).
  • Properties: snake_case, typed, no PII.
  • Server-confirmed actions fire from the backend; UI views fire from web.
  • Reuse existing invoice_sent instead of a new duplicate.
2. Event table
Event nameTriggerPlatformProperties (type, example)Metric it feeds
recurring_setup_viewedSetup modal finishes renderingWebentry_point (string, "invoice_detail")Setup completion rate
recurring_schedule_createdAPI returns 201 for new scheduleBackendschedule_id (string, "rs_8812"), frequency (string, "monthly"), amount_cents (int, 45000)Setup completion rate, Adoption rate
invoice_sentInvoice email accepted by mail providerBackendinvoice_id (string, "inv_2290"), is_recurring (bool, true)Recurring invoices sent per account
recurring_schedule_pausedAPI confirms pauseBackendschedule_id (string), days_active (int, 64)90-day schedule retention
recurring_schedule_cancelledAPI confirms cancelBackendschedule_id (string), cancel_reason (string, "client_ended")90-day schedule retention
3. Metric coverage
  • Setup completion rate = unique accounts with recurring_schedule_created / unique accounts with recurring_setup_viewed
  • Adoption rate = active accounts with ≥1 recurring_schedule_created / all active accounts (needs existing account_active definition)
  • Recurring invoices sent per account = count of invoice_sent where is_recurring = true / adopting accounts
  • 90-day schedule retention = schedules created 90+ days ago with no recurring_schedule_cancelled / schedules created 90+ days ago
4. Cut or justify
  • frequency_dropdown_opened: feeds no metric. Cut unless design wants to test dropdown usability.
  • recurring_tooltip_hovered: feeds no metric. Cut.
5. QA checklist
  1. Each event fires once per action in staging (check for double fires on retry).
  2. recurring_schedule_created does not fire when the API returns an error.
  3. is_recurring is set on every invoice_sent, never null.
  4. No property contains a client name or email.
  5. Amplitude chart for each metric formula returns data within 24 hours of release.

Assumptions: "active account" already exists as a definition; pause is counted as retained, which needs your decision.

Written by AI from this prompt. Your results will vary.

Tips and common edits

Why this works

  • The object_action, past-tense rule with a bad example gives consistent names engineers and analysts can search.
  • Forcing each event to map to a metric pushes orphan events into a visible cut list before they ship.
  • Requiring metric formulas built from events exposes missing instrumentation while it is still cheap to add.

Use before engineering starts instrumenting a new feature. Have your user flow, success metrics and current event names ready. Works in ChatGPT, Claude or Gemini; paste the table into your tracking spec.

Common edits

  • Ask for the event table as CSV so you can import it into your tracking spec or Avo.
  • Add a column for owner and implementation status to use it as a launch checklist.

Prompt details

By
Prompts-GPT editors
Published

Did this prompt work?

Keep exploring

More in Product Management