Alerts that wake the right person, not everyone

There's a failure mode worse than having no alerts: having alerts everyone has learned to ignore. A channel that lights up on every single exception is noise, and noise is invisible. The whole point of monitoring is that when something important happens, the right person notices — once.

This is how Lookout is built to keep alerts quiet enough to be worth reading.

First, grouping turns a flood into an issue

Before anything alerts, Lookout groups events by fingerprint. A thousand copies of the same QueryException from the same place collapse into one issue with a count of a thousand — not a thousand rows and not a thousand notifications. You can still drill into any single occurrence for the exact request and stack trace, but the unit you work with (and alert on) is the issue, not the raw event.

That alone removes most of the noise. The error you need to see isn't buried under its own duplicates.

Then, alert on a threshold — not on every event

An alert rule fires when volume crosses a line you set over a time window, rather than on each event:

• pick a window — 1 hour or 24 hours,
• set a threshold for how many events in that window matter,
• choose a channel — Email or Slack.

A scheduled check evaluates the rule and notifies you when the threshold is crossed. So a sudden spike pages you once; a slow background hum of low-volume noise doesn't page you at all. You're tuning for "tell me when this is actually happening," not "tell me it exists."

A sensible starting point

You don't need many rules. A good default for one project:

• A 24h rule with a modest threshold for "this issue is recurring and someone should look" → Email.
• A 1h rule with a higher threshold for "something is on fire right now" → Slack, where on-call will actually see it fast.

Two rules, two windows, two urgencies. Start there and adjust the thresholds after a week of watching real traffic — the right number is the one that's quiet on a normal day and loud on a bad one.

Why this matters more than detection

Every error tracker can detect errors. The thing that decides whether monitoring actually helps is whether the alerts stay credible. An alert that fires constantly gets a filter rule in someone's inbox and is never seen again. An alert that fires when it means something gets read. Grouping plus thresholds is how you stay on the second side of that line.

Try it on a real project

Grouping and alert rules are available on the free Starter tier — 10,000 events a month, one project, no credit card. Point it at something you ship, set two rules, and let a normal week calibrate your thresholds.

Create your free project and set up an alert that's worth not ignoring.