Find your magic activation moment

Why this matters

The single most important question in early-stage SaaS analytics: what’s the action that, once a user takes it, predicts they’ll stick around?

For Slack it was sending 2,000 messages. For Facebook it was 7 friends in 10 days. For Notion it’s reportedly creating 5 pages. Every successful product has this — and finding it shifts your entire onboarding strategy because suddenly there’s a single bar to clear.

Most teams never answer this question because it requires:

• A 30-day retention definition
• Cohorting users by first-24h behavior
• Running retention per cohort
• Comparing lift

That’s 3 SQL queries and a spreadsheet. Or one prompt to Agentry.

What you get

• The top 3 actions ranked by retention lift
• Conversion + retention curves for each cohort
• A saved cohort definition for “activated users” that the rest of Agentry can use (cohort-based feature flags, surveys, etc.)

Walk through it

agentry_list_event_names {
  project_id: "default",
  min_count: 100
}

{
  "events": [
    "$pageview", "signup_completed", "first_project_created",
    "team_member_invited", "first_query_run", "integration_connected",
    "tutorial_completed", "subscription_started", "comment_added", "..."
  ]
}

agentry_analytics_query {
  project_id: "default",
  hogql: "
    WITH
      cohorts AS (
        SELECT
          person.id AS person_id,
          min(timestamp) FILTER (WHERE event = 'signup_completed') AS signup_ts,
          array_agg(DISTINCT event) FILTER (WHERE timestamp <= signup_ts + INTERVAL 24 HOUR) AS first_day_events
        FROM events
        WHERE event = 'signup_completed' AND timestamp > now() - INTERVAL 60 DAY
        GROUP BY person.id
      ),
      retained AS (
        SELECT person_id, count() AS day_30_events
        FROM events
        WHERE timestamp BETWEEN now() - INTERVAL 31 DAY AND now() - INTERVAL 28 DAY
        GROUP BY person_id
      )
    SELECT
      arrayJoin(first_day_events) AS first_day_event,
      count(DISTINCT cohorts.person_id) AS cohort_size,
      countIf(retained.day_30_events > 0) AS retained_count,
      retained_count / cohort_size AS retention_rate
    FROM cohorts LEFT JOIN retained USING (person_id)
    GROUP BY first_day_event
    HAVING cohort_size >= 50
    ORDER BY retention_rate DESC
    LIMIT 10
  "
}

{
  "rows": [
    { "first_day_event": "integration_connected", "cohort_size": 142, "retention_rate": 0.78 },
    { "first_day_event": "team_member_invited", "cohort_size": 98, "retention_rate": 0.71 },
    { "first_day_event": "first_query_run", "cohort_size": 320, "retention_rate": 0.64 },
    { "first_day_event": "first_project_created", "cohort_size": 480, "retention_rate": 0.51 },
    { "first_day_event": "$pageview", "cohort_size": 620, "retention_rate": 0.18 },
    { "first_day_event": "tutorial_completed", "cohort_size": 84, "retention_rate": 0.16 }
  ]
}

The output

🎯 Magic activation moment found

Top action by retention lift (first-24h → day-30 active):

  1. integration_connected     78%  (4.3x baseline)
  2. team_member_invited       71%  (3.9x baseline)
  3. first_query_run           64%  (3.5x baseline)

Negative signal:
  tutorial_completed           16%  (below baseline 18%)
  → tutorial is replacing real engagement, not feeding it

Saved cohort: `activated_users` = "connected an integration within 24h of signup"
  Use it in: agentry_create_feature_flag, agentry_create_survey,
            agentry_analytics_query (`WHERE cohort = 'activated_users'`)

Next moves:
  • Onboarding: make integration setup the primary CTA on day 1
  • Tutorial: A/B test removing it for new signups
  • Email: 24h-later nudge to integrate, with one-click suggestions

Setting it up

You need two kinds of events:

• A signup_completed event (or equivalent — the agent matches by intent)
• Other events captured in the first 24h that are candidates for “the magic action”

Both should have distinct_id set so cohorts work. Agentry has no SDK — everything is a raw POST to /v1/analytics/. The tiny sendAnalytics helper below is the kind of thing your agent generates at install time (~25 lines, edit freely):

// Agentry.ts — generated by your agent at install time. NOT an SDK.
const DSN = process.env.AGENTRY_DSN!;
const PROJECT_ID = process.env.AGENTRY_PROJECT_ID!;
const UA = "myapp/1.0";  // REQUIRED — Cloudflare 403s default UAs

export async function sendAnalytics(body: unknown) {
  await fetch(`https://api.agentry.sh/v1/analytics/${PROJECT_ID}/`, {
    method: "POST",
    headers: {
      "Authorization": `Bearer ${DSN}`,
      "Content-Type": "application/json",
      "User-Agent": UA,
    },
    body: JSON.stringify(body),
  });
}

// Signup completion — set user properties via $set on the same event
await sendAnalytics({
  event: "signup_completed",
  distinct_id: user.email,
  properties: { $set: { signup_date: new Date().toISOString() } },
});

// Candidate activation events — instrument generously, the agent figures
// out which one matters
await sendAnalytics({
  event: "integration_connected",
  distinct_id: user.email,
  properties: { provider: "github" },
});
await sendAnalytics({
  event: "team_member_invited",
  distinct_id: user.email,
  properties: { count: 1 },
});
await sendAnalytics({
  event: "first_query_run",
  distinct_id: user.email,
  properties: { query_id: q.id },
});
await sendAnalytics({
  event: "first_project_created",
  distinct_id: user.email,
  properties: { project_id: p.id },
});

The more candidate events you capture, the better the agent’s analysis. Don’t worry about overcapturing — Agentry’s pricing is reasonable on volume, and unused events cost nothing to query later.

Variations

• “Same analysis but split by paid vs free signups — does the activation moment differ?”
• “What’s the activation moment for users who came from organic search vs paid ads?”
• “Look at the 90-day retention version — does the same moment hold up over a longer window?”
• “For users who hit the magic moment, what was the most common SECOND action they took? That’s the day-2 nudge target.”