Back to Cookbook

Freshness Alarm

Know about stale data before your stakeholders do

Monitors your key tables for freshness, row counts, and schema changes on a schedule. Alerts you via Slack or Telegram before business hours when something looks off — so you're never blindsided by a stakeholder asking "why is the dashboard blank?"

CommunitySubmitted by CommunityWork5 min setup

INGREDIENTS

💬Slack✈️Telegram

PROMPT

Create a skill called "Freshness Alarm". I'll give you a list of database tables with their expected update frequency (e.g., "orders table should update daily", "events table should update hourly"). On a schedule I define, check each table for: (1) Freshness — when was the most recent row inserted or updated? Is it within the expected window? (2) Row count — is today's count within the normal range based on the last 30 days? (3) Schema — have any columns been added, removed, renamed, or changed type since the last check? (4) Null rates — have any columns suddenly become mostly NULL? For each issue found, send an alert to my configured channel (Slack or Telegram) with severity level (critical/warning/info), the specific issue, and suggested next steps. Generate a daily summary of all checks.

How It Works

The most embarrassing moment in analytics is when the VP asks why the dashboard

numbers look wrong and you have no idea the pipeline failed three days ago.

This skill runs checks before your stakeholders wake up and alerts you when

something needs attention.

What You Get

  • Scheduled freshness checks on key tables (configurable frequency)
  • Row count monitoring with anomaly detection (sudden drops or spikes)
  • Schema change detection (new/removed/renamed columns, type changes)
  • Null rate monitoring per column
  • Alerts via Slack or Telegram with severity levels
  • A daily health dashboard summarizing the state of all monitored tables

Setup Steps

  1. Ask your Claw to create a "Freshness Alarm" skill with the prompt below
  2. List the tables to monitor and their expected update frequency
  3. Configure alert channels (Slack, Telegram, email)
  4. Set the check schedule (e.g., 6am before business hours)
  5. Optionally set row count thresholds and null rate baselines

Tips

  • Start with the tables behind your most critical dashboards
  • The schema change detection catches the vendor-side changes that break pipelines
  • Row count anomalies often catch pipeline failures faster than freshness alone
  • Pair with Pipeline Medic for automatic root cause analysis when an alert fires
Tags:#data-quality#monitoring#pipelines#alerting