# Move from live traffic to measurable production behavior.

> Explore production events and latency percentiles with service, route, method, trace, geography, source, status, and error filters.

TrafficWar keeps live investigation and historical event analysis connected: start with a traffic change, then narrow it to the exact service, route, status, or trace.

## Percentiles instead of averages alone

Metrics include counts, error counts, rates, moments, and latency percentiles. The mergeable histogram model keeps p50 through tail percentiles available across live and retained metric windows.

- Group by event category, route label, or service.
- Compare HTTP methods and status families.
- Continue live metrics from a loaded historical baseline.
- Export service-scoped hot metrics in Prometheus exposition format.

## Raw events when details matter

Explore exposes the event fields needed for diagnosis, including event time, received time, trace ID, operation type, source, geography, status code, canonical error state, and tenant-owned properties.

## Retention follows the plan

Live inspection is available on every plan. Historical raw coverage can retain all events, failures and traces, or metrics only. The pricing page shows the current limits returned by the public plan API.

## Related

- [/pricing](https://trafficwar.tech/pricing.md)
- [Use stable event identities that stay useful at scale.](https://trafficwar.tech/docs/event-schema.md)
- [Query service-scoped events programmatically.](https://trafficwar.tech/docs/query-api.md)