For AI agents: a documentation index is available at /docs/llms.txt. Append .md to any page URL for markdown, or send Accept: text/markdown.
No activation event: choose and validate one
Activation is the first moment a new user clearly gets value from your product. To measure activation, you need an event that marks that moment. Many teams lack an activation event or use a proxy that doesn't represent value, such as a page view. Without the right event, activation and early retention analysis use an inaccurate signal.
Data Assistant can't choose an activation event for you. Activation is a business decision about your product, and Data Assistant can't detect that the event data needed to measure activation is missing. Choosing and validating an activation event requires judgment, but most of the work requires no code.
Choose and validate an existing event before you ask anyone to write code. Use Funnel Analysis and Retention Analysis after you find a candidate. Compass can surface behaviors that correlate with retention. A correlation doesn't prove that a behavior causes retention.
Choose the activation outcome
- Describe the value the user receives, not the screen or button they select. Prefer a completed collaboration over a page view, for example.
- Choose the earliest outcome that reflects real value and that your team can influence.
- In Data > Events, search for an event whose definition matches that outcome.
- If a candidate exists, open the event details and confirm the trigger, source, owner, and properties.
- Build a Funnel Analysis chart from a new-user starting event to the candidate event.
- Build a Retention Analysis chart or a Compass chart to check whether users who reach the candidate return or convert at a different rate.
- Review the result with product, data, and engineering owners before you adopt the event as the activation signal.
Amplitude options and limitations
If one existing event already describes the full activation outcome, use the event directly. When any of several existing events qualifies, a custom event combines the events with an OR condition. The custom event can serve as the analysis signal without an application release. A custom event doesn't create raw telemetry or combine an ordered sequence of actions into a new event.
If the outcome isn't captured anywhere, no in-product tool can invent the signal. In that case, you need new instrumentation.
Ask Global Agent
Global Agent can find relevant existing events and build analyses that compare them. Try prompts like:
"Is there an event that represents when a new user first gets value from the product?"
"For users who signed up in the last 90 days, which early actions correlate with staying active after four weeks?"
Review the generated analysis and supporting data before you choose a signal. Global Agent can't create a signal you never tracked.
Send requirements to the tracking owner
When you need a new signal, send your engineers a precise definition. Copy this and fill in the brackets:
We need an activation event that fires when a new user reaches [OUTCOME, in
product terms, not a screen]. Trigger: [WHEN EXACTLY]. It should carry [USER ID +
these properties]. No existing event represents this outcome. Add it to
the tracking plan and instrument it. We need it for activation and retention
analysis.
If you have no engineering resources
If a custom event can't cover the outcome, you need a source change. Contact the following in order:
- Your Amplitude account team or Customer Success Manager.
- A partner or agency that manages your website or app analytics.
- Amplitude Support at gethelp.amplitude.com for product questions.
When you need to instrument a new activation signal, continue with the engineering solution.
Was this helpful?