Why Your PostHog Events Stopped Working (And How to Fix It)
Your analytics suddenly show zero events. Here's why and how to prevent it from happening again.
You wake up, check your dashboard, and see a flat line where your event data used to be. Zero events. Zero users. Zero everything. Your analytics just... stopped.
The 5 Reasons Events Stop
1. API Key Expired or Was Deleted
PostHog personal API keys (phc_) are tied to your user account. If you left the company, were removed from the organization, or the key was manually deleted - everything stops. No warning. No notification. Just silence.
2. You're Over Your Event Limit
Free tier: 1M events/month. Startup: 10M events/month. If you cross that limit, PostHog doesn't stop collecting - they just stop processing. Your events are technically "received" but never show up in your dashboards.
3. SDK Configuration Broke
A deploy that changed your PUBLIC_KEY, a CDN caching issue, or a typo in the project ID can silently break tracking. The SDK won't throw errors - it just quietly fails to send anything.
4. Project Got Deleted
PostHog allows deleting projects. If someone on your team accidentally deleted the project (it requires typing the project name to confirm), all events sent to that project vanish.
Pro Tip: Use Project API Keys (phg_)
Project-scoped keys (phg_) don't get invalidated when a user leaves. They're tied to the project, not a person. Use these for client-side tracking and avoid the "user left = tracking broke" problem.
How to Prevent Event Blackouts
1. Set Up Event Volume Alerts
Configure alerts for when events drop below 50% of your 7-day average. This catches issues within hours, not days.
2. Create a Health Check Endpoint
Add a simple check that fires every minute: send a test event and verify it appears in PostHog within 2 minutes. If it doesn't, alert your team.
3. Use Project Keys for Client-Side
Don't use personal API keys for tracking. Use project API keys (phg_) which are tied to the project, not a user account.
The Bottom Line
PostHog event blackouts are more common than you'd think. The solution isn't to panic - it's to have monitoring in place BEFORE the flatline appears. A simple "events less than 50% of average" alert would have saved all the companies I talked to this month.