Detection time is half your outage
Total incident duration is detection plus response plus recovery. Teams pour enormous effort into the last two and then leave a five-minute check interval in place, which sets a floor on the first that no amount of on-call excellence can overcome.
If your checks run every five minutes, your average outage is already two and a half minutes old before anyone could possibly know. Moving to sixty seconds removes two minutes from every single incident you will ever have.
Not every endpoint deserves the same interval
- Revenue pathsCheckout, payment callbacks, and login. 30 seconds or faster — the cost of downtime dwarfs the cost of checking.
- Core APIsAnything other services depend on. 60 seconds is a sensible default.
- Marketing pagesImportant but not transactional. 5 minutes is usually plenty.
- Batch and internal toolsLow urgency. 15 to 30 minutes avoids noise for things nobody uses at 3am.
Multiply interval by monitor count by region count and the numbers grow quickly. Ten monitors at 30-second intervals across five regions is 100 requests a minute against your own infrastructure, continuously, forever. Make sure the endpoints you point at are cheap to serve.
Prefer cheap endpoints for frequent checks
A health endpoint that runs three database queries and renders a template is a bad monitoring target at high frequency — you are adding meaningful load precisely when the system is already struggling.
The better pattern is a lightweight endpoint that verifies critical dependencies with minimal work, returns a small response, and is excluded from caching. Save the expensive, realistic transaction checks for a slower interval.