Daily Metrics Monitor is a Claude scheduled task (skill) that works as an autonomous daily growth analyst for a subscription business. Every morning it pulls 30+ metrics across all revenue funnels — web subscription funnel, iOS app, Android app — from Campaignswell, Stripe, RevenueCat and PostHog, checks every number against red/yellow flag thresholds, investigates anything that fired, publishes a formatted report page to a Notion database, and sends a one-line Slack summary. The whole run takes under 8 minutes with zero human involvement.
Pulls the full metric picture daily. Conversion rates per funnel and platform, purchases, spend, predicted ROAS with confidence intervals, predicted 6-month LTV by weekly cohort and country tier, renewal rates by billing cycle and plan, churn at 1/3/6 months, upsell revenue, subscription mix, MRR, active subscribers. Every metric comes with today / yesterday / 7d / 30d windows plus 4 weekly cohorts and week-over-week deltas.
Watches billing health at the raw level. Straight from the payment processor: refunds, disputes, fraud-risk transactions, payment approval rates split into initial purchases vs renewals, and failed payments grouped two ways — the processor’s interpretation AND the raw bank decline code. The raw code is what tells you whether it’s insufficient funds or a bank suddenly blocking you.
Flags problems against explicit thresholds. LTV below floor, refund rate above ceiling, any major metric down >15% WoW, a new decline code entering the top 3, prepaid-card share spiking, a funnel step degrading >10%, a creative burning budget below break-even ROAS.
Investigates every flag automatically. Each flag has a playbook: an LTV drop triggers a country-tier split, a renewal-curve breakdown by billing cycle and a subscription-mix check; a CR drop triggers a step-by-step funnel analysis with a locale drill-down. The report doesn’t say “LTV is down” — it says why, with evidence and one concrete next step.
Degrades gracefully. If a data connector fails, the report still ships with what’s available, marked Partial, with failed connectors listed. The Notion database itself becomes the fallback source of history for computing deltas.
The obvious value is time: it replaces roughly 60 minutes of daily dashboard scrolling across four tools with an 8-minute autonomous run — about 5 hours a week, 20+ hours a month of PM/analyst time returned.
The bigger value is detection latency. The dangerous failure mode of manual monitoring isn’t the hour a day — it’s the days you skip the check and a refund spike, a broken payment flow, or a traffic-quality shift runs unnoticed for a week. With the monitor, incidents surface the next morning with a diagnosis attached, not a week later as a revenue hole.
A third effect: the report database becomes your metrics system of record. Every day is one page with identical structure and properties, so “when did refund rate start climbing” is answered by scrolling one database, not re-querying four tools.
MCP connectors for your stack (the skill is written for Campaignswell + Stripe + RevenueCat + PostHog + Notion + Slack, with 🔁 SWAP markers showing exactly what to replace for other BI, payment or analytics platforms), a scheduled-task runner, and one setup afternoon: verify your platform’s exact metric/column names once and hardcode them, set your own LTV floors and refund ceilings, point it at your Notion database.
The single most important setup rule, learned the hard way: treat the prompt like production config. Exact column names instead of “find the LTV metric”, explicit thresholds instead of “flag anything unusual”, defined failure behavior instead of hoping. Every ambiguity left in the prompt is a metric that comes back wrong tomorrow at 9:30.
You are the company’s daily growth analyst. Your job: pull all metrics across funnels, save a richly-formatted Notion page, and ping the owner with one line in Slack. You replace ~60 minutes of dashboard scrolling.
IMPORTANT LESSON FROM EARLIER VERSIONS: if the prompt says “search for metric X”, half the metrics come back missing. This version uses pre-verified column names from your analytics platform — verify them once during setup, then hardcode them below. DO NOT search for column names at runtime.
═══════════════════════════════════════════ SETUP — replace before first run ═══════════════════════════════════════════
<CW_ORG_ID> — your Campaignswell organization ID