Beginner mistake I made: Not setting up proper filters. Bill shock story.

Ouch, that's a painful way to learn that lesson, but thank you for sharing it so clearly. Your point about filtering *before* ingestion is crucial; it's so easy to think you'll just clean the data later, but the meter is already running.

One thing I'd add from moderating similar stories: it's also wise to set up billing alerts if the platform offers them, as a secondary safety net. Sometimes a misconfigured filter or a new, unexpected traffic source can still slip through. That alert gives you a chance to react before the next weekly or monthly tally.

Great point about the alerts. We had a filter fail silently after an API update, and the alert saved us. It's a simple belt-and-suspenders move.

Just make sure the alert goes to a channel someone actually watches, not a dead email inbox. In our case, we route it to a dedicated Slack channel for monitoring alerts.

Your list is spot on. Add one more: filter by HTTP method. We got torched on a monitoring system that was polling via POST on an endpoint we'd only filtered GET for. It's easy to forget that internal systems don't always follow the same patterns as client traffic.

Also, "thousands of traces from automated systems" - you can quantify that with a quick tag breakdown in the tracing UI. Show your team the percentage of traces that are noise. It turns a lesson into a concrete metric for justifying the time spent.