Why your engineering Slack channel is so noisy (and how to fix it)
The main culprit is bot integrations posting per-event notifications. GitHub alone can dump 100-300 messages per week into a channel. Fix it by switching to scheduled digests with filters, separating channels by purpose, and auditing which integrations post the most.
What causes most Slack noise in engineering teams?
Git notifications are usually the biggest source, sometimes over half the messages in a channel. After that: CI/CD alerts, monitoring/incident pings, deploy notifications, and bot status updates. Most of these don't need to be per-event.
How do I audit Slack notification usage?
Check Slack analytics for messages per channel per week. Flag channels over 100 messages/week where bots outnumber humans. Then check which integrations post the most. GitHub, GitLab, and CI tools are usually the worst offenders.
Should we use separate channels for different things?
Generally yes. Split out PR notifications, CI/CD, incidents, deploys, and general discussion. Then use a digest tool like HotBot to keep the PR notification channel to one message per day. Keeps signal high in each channel.
How do notification filters work?
Each PR gets checked against rules you define: skip drafts, skip anything labeled 'wip', skip bot-generated PRs (dependabot, renovate), skip PRs already reviewed, skip PRs touching certain file paths. Only what passes the filters shows up in the digest.
What happens after we reduce the noise?
Teams that switch to digests tend to get faster at reviewing PRs (less scrolling to find them), show up to standup more prepared (the digest lands beforehand), and generally complain less about Slack being unusable. Message volume drops a lot.