Activation funnels

Treat activation as a funnel you can query, not a vibe. Every new user flows through the first-run path, so effects show up quickly — which also makes it easy to optimize the wrong proxy.

Use product analytics so each step is a named event with stable properties (template_id, source=organic|paid, persona). Break funnels by acquisition channel: paid traffic often activates worse than organic.

Walking through a leak

A typical skeleton:

Signed up
  → Completed account essentials (email verify / workspace name)
    → Reached first-value screen (template chosen / data connected)
      → Did activation event (invite / first query / first order)
        → Repeated at natural frequency within D days

8,000 weekly signups. 72% verify email. 40% reach the template picker. 22% create a project from a template. 11% invite a teammate within 7 days (the activation event). Week-4 retention: 48% for activated vs 9% for the rest.

The biggest absolute leak is template picker → project (40% → 22%). That is where you experiment first — not the invite modal.

Key metrics:

• Activation rate: % of signups that hit the activation event within the window (e.g. 7 days).
• Time-to-activation: median / p90 hours to that event.
• Step conversion: drop-off between consecutive steps.
• Activated vs unactivated retention: the sanity check.

A minimal event plan:

signed_up(user_id, channel, persona)
onboarding_step_viewed(step_id)
onboarding_step_completed(step_id)
template_selected(template_id)
integration_connected(provider)
activation_event(definition_version)
first_return_session(days_since_signup)

Version the activation definition and store the window (e.g. 7 days) in the metric definition. When you change "aha" from "invite" to "first shared edit," old charts must not silently lie.

Onboarding is a high-ROI surface for A/B tests. Guardrails matter: do not declare victory on checklist completion if retention is flat. Segment paid vs organic before rewriting the whole flow.

In interviews, "how would you improve onboarding for X?" is a classic. Structure: define the user and job, propose an activation metric, map the funnel, pick the biggest leak, design one experiment with a retention guardrail. Mention how you would validate the definition before optimizing it.

Common mistakes

• Optimizing the last step because it sounds like the aha. Fix the largest absolute drop first.
• Blending channels. Paid users often need more hand-holding. If you blend them, you over-educate high-intent users or under-serve low-intent ones.
• No definition_version. Changing the aha rewrites history.

Check yourself

Funnel: 40% reach templates, 22% create a project, 11% invite. Where do you experiment first?

Template → project. That drop (18 points) is the largest leak.

Activation rate doubles. Week-4 retention is unchanged. What now?

The proxy is not causal. Retire it and find an event that still predicts retention.

Key takeaways

• Instrument signup → first value → aha → return, and segment by channel.
• Experiment on the biggest drop-off, with retention as the guardrail.
• Version the activation definition so charts stay honest.

Further learning

Back to the map in activation and onboarding. Pair with A/B testing on the leak you just found.