An alert checked every 15 minutes runs 96 times a day and should mail you once. How rising-edge suppression, cooldowns and recovery mails keep alerts read.
An alert checked every 15 minutes runs 96 times a day. It should mail you once.
An alert that mails every time it is true is a newsletter. By the second day, nobody reads it.
Our Customer-Demo sample project ships four alert rules. Each one asks a different question:
📉 Daily revenue below plan: a SQL query on the project's DuckDB, fires when SUM(revenue) falls under 150,000, goes to sales leads and finance with the table attached.
🚨 Single order over the limit: fires when any one row is over 25,000, and puts the breaching rows in the e-mail for the risk team.
📅 New month in the sales feed: watches a pipeline's data without running it. Its first check only notes the month already there; the next new month sends.
📋 Board pack numbers changed: watches a range on a sheet. Switched off, configuration kept, ready for the next board cycle.
What keeps them readable is what happens after the condition is true:
Before a rule runs alone, Check now tries it this moment, suppression included, even while it is switched off. A missing source or a rejected send shows up as a failed run in the scheduler, never as an alert that quietly did not fire. Rules are saved in the project and versioned in Git.
Since 1.6 an alert can add a short AI paragraph to its e-mail. The rule decides whether it fires, never the AI.
One limit, stated plainly: e-mail is the only channel today. No Teams, Slack or webhook yet.
Handbook chapter and download: links in the first comment.
What is the one alert your team has learned to ignore?
This site needs JavaScript for the full experience. Every page above is reachable without it.