Telemetry
Browse docs
GuidesUpdated July 28, 2026Reviewed by the Telemetry product team2 min read

Use this doc with your coding agent

Copy a Telemetry prompt and ask Claude Code, Codex, or Cursor to instrument the workflow covered here.

PromptsGet API key
On this page
  1. Emit durable milestones
  2. Calculate two forms of conversion
  3. Segment only after the base journey works

Product Journey Analysis with SQL

A product journey can count sessions, users, accounts, or attempts. Choosing the unit after looking at the result makes the analysis unstable. Define the counting unit, ordered milestones, identity rules, and observation window before writing SQL.

Emit durable milestones

Use event names that describe completed product outcomes, such as landing_viewed, signup_completed, workspace_created, and first_query_run. Include a privacy-reviewed session or account identifier, a stable step name, step order, bounded acquisition source, and UTC timestamp.

Do not infer a completed milestone from a button click when the underlying operation can fail. Deduplicate repeated events when the producer can retry. Document whether the journey must occur within one session and whether users may skip or repeat steps.

Calculate two forms of conversion

The product journey SQL recipe produces:

  • journey_completion_pct, which compares each step with the first step.
  • step_to_step_pct, which compares each step with the immediately preceding step.

The first answers how much of the entering population reaches a milestone. The second locates the transition with the largest local loss. The included fixture has five entering sessions and two that reach the first-query milestone, so both percentages can be reproduced in the browser SQL playground.

For stricter production funnels, first reduce to one earliest milestone per session, verify step order with window functions, and choose a maximum journey duration. The SQL funnel analysis guide explains those choices in more detail.

Segment only after the base journey works

Validate the unsegmented journey before adding source, plan, device, or experiment dimensions. Each segment needs enough complete observation time and enough sessions to support a decision. Keep raw count beside every rate.

The product analytics use case connects journey SQL with activation and retention. Use the feature rollout guide when the question is whether a controlled release changed the journey or its reliability.