Find your top 10 power users — and email them

Why this matters

Paul Graham’s “do things that don’t scale” essentially boils down to: talk to your best users. The 10 people who use your product more than anyone else are a goldmine — they tell you what to build, they refer their friends, they renew without thinking.

The only thing stopping every SaaS founder from doing this weekly is friction. “Who are my top users?” is a SQL query against your events table. Most founders are 6 dashboards deep and never quite pull the trigger.

Agentry does it in one prompt because the agent already knows your event vocabulary, can weight by intent (revenue > vanity > pageviews), and can output it as a paste-able list.

What you get

• A ranked list of your top users by weighted activity
• Each user’s email, signup date, plan, and top 3 actions
• A copy-paste-ready Markdown table for your weekly review doc

Walk through it

agentry_list_event_names {
  project_id: "default",
  matching: "subscription|payment|invite|share|comment|created"
}

{
  "events": [
    { "name": "subscription_started", "count": 142 },
    { "name": "subscription_upgraded", "count": 38 },
    { "name": "invoice_paid", "count": 480 },
    { "name": "team_member_invited", "count": 86 },
    { "name": "comment_added", "count": 1240 },
    { "name": "project_shared", "count": 91 },
    { "name": "project_created", "count": 720 }
  ]
}

agentry_analytics_query {
  project_id: "default",
  hogql: "
    SELECT
      person.properties.email AS email,
      person.properties.signup_date AS signup_date,
      person.properties.plan AS plan,
      sum(CASE WHEN event IN ('subscription_started','subscription_upgraded','invoice_paid') THEN 3
               WHEN event IN ('team_member_invited','comment_added','project_shared') THEN 2
               ELSE 1 END) AS weighted_score,
      arrayJoin(topK(3)(event)) AS top_action,
      count() AS total_events
    FROM events
    WHERE timestamp > now() - INTERVAL 30 DAY
      AND person.properties.email IS NOT NULL
    GROUP BY email, signup_date, plan
    ORDER BY weighted_score DESC
    LIMIT 10
  "
}

| # | Email                  | Plan  | Signup     | Score | Top Actions                          |
|---|------------------------|-------|------------|-------|--------------------------------------|
| 1 | [email protected]         | pro   | 2026-01-12 | 487   | project_created, comment_added, …    |
| 2 | [email protected]     | scale | 2025-09-04 | 412   | invoice_paid, project_shared, …      |
| 3 | [email protected]        | pro   | 2026-02-28 | 388   | comment_added, project_created, …    |
| 4 | [email protected]     | free  | 2026-04-19 | 364   | project_created, $pageview, …        |
| 5 | [email protected]     | pro   | 2025-11-22 | 351   | team_member_invited, comment_added, … |
…

Two things jump out:

• #4 [email protected] is on FREE plan but has more weighted activity than 80% of pro users. Strong upgrade candidate. Personal email from you would probably convert them.
• #2 [email protected] has high invoice_paid count (likely on monthly billing) — worth a check-in to ask about annual.

Want me to draft personal emails for the upgrade candidates?

The output

🏆 Top 10 power users (last 30 days, weighted)

  1. [email protected]         pro    score 487  💪 active daily
  2. [email protected]     scale  score 412  💰 monthly billing — pitch annual
  3. [email protected]        pro    score 388
  4. [email protected]     free   score 364  ⚠️ upgrade candidate
  5. [email protected]     pro    score 351
  … (5 more)

Actionable callouts:
  • 1 free user with pro-tier activity → personal upgrade email
  • 1 monthly customer with consistent usage → annual upsell

Generated table is paste-ready for your weekly review doc.

Setting it up

Agentry has no SDK and no separate identify() call. Every analytics event is a raw POST to /v1/analytics/, and you attach the identifier + user properties inline in the event body.

1. distinct_id set to the user email (or any stable per-user identifier) on every event:

await fetch(`https://api.agentry.sh/v1/analytics/${process.env.AGENTRY_PROJECT_ID}/`, {
  method: "POST",
  headers: {
    "Authorization": `Bearer ${process.env.AGENTRY_DSN}`,
    "Content-Type": "application/json",
    "User-Agent": "myapp/1.0",  // REQUIRED — Cloudflare 403s default UAs
  },
  body: JSON.stringify({
    event: "project_created",
    distinct_id: user.email,
    properties: { project_id: project.id },
  }),
});

2. User properties via $set on any event so the agent can rank by plan / signup date. PostHog (which /v1/analytics/ forwards to) reads $set and updates the person record:

await fetch(`https://api.agentry.sh/v1/analytics/${process.env.AGENTRY_PROJECT_ID}/`, {
  method: "POST",
  headers: {
    "Authorization": `Bearer ${process.env.AGENTRY_DSN}`,
    "Content-Type": "application/json",
    "User-Agent": "myapp/1.0",
  },
  body: JSON.stringify({
    event: "$identify",       // or piggyback $set on any event you already send
    distinct_id: user.email,
    properties: {
      $set: {
        plan: user.plan,           // "free" | "pro" | "scale"
        signup_date: user.created_at,
        team_id: user.team_id,
      },
    },
  }),
});

That’s it. The agent figures out the rest from your event names.

Variations

• “Show me the top 10 free-tier users. I want to email them personally about upgrading.”
• “Find users who were top-10 last month but dropped out of the top 50 this month — they’re cooling off.”
• “Power users by team_id instead of email — which companies are the most engaged?”
• “For each top user, draft a 2-sentence email asking them what they wish Agentry could do.”