Event spec

What to measure at each stage, and the events that feed it.

The rule

Never a count on its own. Every volume metric sits next to a quality metric, so a spike from a Hacker News post can't be mistaken for growth. Watch trends, not absolute numbers. The headline metric is time to first signal.

Five stages

Stage	Volume metric	Quality metric
Awareness	Docs visits	Depth: pages on scaling, security, retention, RBAC
Acquisition	Signups	Signup within an hour of first visit
Activation	setup_success count	Time from signup to setup_success
Retention	Daily active workspaces	Workspaces active 7 of the last 14 days
Commercial trigger	Accounts hitting a paid gate	Accounts that hit a gate and were touched within 24 hours

The six events

These are the minimum. Nothing on the dashboard is true until they exist.

setup_started          user began the install or Cloud setup flow
setup_success          first telemetry accepted
auth_attempt           user tried to log in to a workspace
api_key_created        user generated an API key
docs_page_path_depth   which docs section, how many pages in
copy_code_sample       a code block was copied, and which one

Conversion chain

Impressions, then docs visits, then signups, then first successful call, then daily use. Each arrow has a number. The number that matters most is how long the third arrow takes.

Time to first signal

Benchmark for infra tools is 45 to 90 minutes from signup to first data flowing. Above that, most users don't come back. Most activations happen inside the first hour or not at all.

What I'd look at first

• Time to first signal, weekly, as a distribution rather than an average
• Which docs page people were on right before setup_success, and right before they stopped
• Which copy_code_sample blocks show up in accounts that activate, and which show up in accounts that don't
• Percentage of events arriving untagged. If that number is above a few percent, nothing else on the board can be trusted.

Movements, not totals

Report new, expansion, contraction, churn, and reactivation as separate lines. A flat total can hide a business where new logos have stopped and churn has improved, and those two things need opposite responses.

Assumptions

Marked as assumptions until the data says otherwise:

• Landing page decides signup rate. Migration and comparison pages convert differently from the homepage.
• Most activations happen in the first hour or never.
• One board per product line. Cloud, self-hosted, Enterprise. Self-hosted has no product events, so its board runs on external signals only.

The funnel dashboard is at gtm-intent-funnel.vercel.app. Layout follows Leah Tharin's PLG template for Amplitude.

---

Source: signoz_assignment_progress.md, Q2 section and product assumptions.