Business Ops
Funnel Watch
A daily watch on your signup and conversion funnel that alerts on real drops with the evidence and stays quiet when healthy.
Moderate setupShort cycleOpen loopCompounds over weeks
How a run flows
Before first run
Verify funnel read access
On schedule
Every day · 8am
Step 1 · Compute
Step-to-step rates
Step 2 · Compare
Against the baseline
Step 3 · Alert
Alert with evidence
Metric
Signup → paid %
Report
Latest check
open loop — repeats on the next tick, picking up where it left off
Prerequisites
- Access to your DB / product analyticsrequired
Read-only access to the funnel data. Manage agent credentials with treg
How this loop works — schedule, exit, notifications
- Suggested schedule
- Daily · ~8am
- Loop type
- Open loop — an ongoing monitor with no finish line.
- Exit condition
- Runs indefinitely (no finish line).
- Notifications
- On by default — you're alerted on each run's outcome (completion or failure). Silence with notify: never.
- Ease of starting
- Some setup and context are needed before the first useful run.
- When you see value
- Quiet by design; its worth shows the day it catches a real conversion drop with the evidence.
The prompt this runs
Set up a daily loop that watches this product's signup and conversion funnel. Before creating it, ask where the funnel data lives - product analytics like PostHog, an events table, an analytics API - verify read access on this machine, run one query as a smoke test, and confirm with me the funnel steps and the conversion rate that matters; never create a blind loop. Every morning around 8am my local time, pull each funnel step for the latest window, compute the step-to-step and overall conversion rates, and compare them against the recent baseline. Alert only on a real drop - a step that fell meaningfully beyond normal daily variance - and when you do, bring the evidence: which step, how far it fell, and the most likely suspects to check (a recent deploy, a broken step, a traffic-source shift). When the funnel is healthy, stay quiet - a short log line, no alarm. Read-only always: never mutate the analytics source, and never copy credentials or personal data into a report. Write a dated report with front matter `type: report`, a `title`, and `date` on any run that flags something, and report the conversion rate as a metric. Set up the loop's dashboard at creation: an embed of the newest report plus a chart of the conversion rate over time.
This is the exact intent appended to the bootstrap when you create the loop — your coding agent reads it, then proposes cadence and config and confirms with you before creating anything.