# See every request, operation, and field across your federated graph

Request rate, P95 latency, error rate, individual traces, operation inventory, schema field usage, and client identification. All built into Cosmo Studio.

Metrics, traces, field usage, and client data. No additional infrastructure required.

## Overview

## Request analytics, built into Cosmo

Cosmo Analytics is the data layer of Cosmo Studio. It collects telemetry from the Cosmo Router via OpenTelemetry and surfaces it across focused views: a high-level metrics dashboard, a request trace list, an operations inventory, schema field usage tracking, and client identification.

All views share the same underlying data and consistent filtering. Filter by client and the selection applies everywhere. Navigate from an operation to its traces without losing context.

## Why GraphQL-aware analytics matter

### Why teams need federation-native analytics

Generic HTTP metrics tell you a request happened. They do not tell you which operation ran, which client sent it, which fields it touched, or which subgraph caused a slowdown. In a federated graph, that context is what matters.

Four gaps that come up repeatedly when teams run federated GraphQL without purpose-built analytics.

### Metrics are disconnected from operations.

HTTP-level monitoring shows request counts and latency. It cannot tell you which GraphQL operation is slow, which client sends it, or how often it runs.

### Schema changes are guesswork.

Teams do not know which fields are used, which clients depend on them, or when it is safe to remove deprecated fields. Changes break things they were not aware of.

### Client traffic is opaque.

When multiple applications consume the same API, anonymous traffic makes it impossible to isolate a client-specific issue or target communication about breaking changes.

### Incident investigation spans too many tools.

Finding the root cause of an elevated error rate means jumping between logs, metrics dashboards, and trace backends — none of which share GraphQL operation context.

Cosmo Analytics handles all of this natively. One data source, six views, consistent filters throughout.

## Cosmo Analytics capabilities

01 Overview & metrics  02 Request inspection  03 Schema intelligence

## Which analytics capability do you need?

| If you are… | Start here |
| --- | --- |
| Checking the health of your federated graph right now | [Metrics Analytics](/content/cosmo/analytics/metrics-analytics/index.html) |
| Debugging a specific failing or slow request | [Trace Analytics](/content/cosmo/analytics/trace-analytics/index.html) |
| Finding which operations have the highest latency or error rate | [Operations Tracking](/content/cosmo/analytics/operations-tracking/index.html) |
| Deciding whether it is safe to deprecate or remove a field | [Schema Field Usage](/content/cosmo/analytics/schema-field-usage/index.html) |
| Filtering analytics data by a specific client application | [Client Identification](/content/cosmo/analytics/client-identification/index.html) |
| Getting a unified view of all traffic with grouping and filtering | [Analytics Dashboard](/content/cosmo/analytics/analytics-dashboard/index.html) |

## How Cosmo Analytics compares

|  | Cosmo Analytics | Generic APM tools | Custom dashboards |
| --- | --- | --- | --- |
| GraphQL operation context | Native | No | Requires instrumentation |
| Schema field usage tracking | Built-in | No | Requires custom build |
| Client identification | Via HTTP headers | Varies | Requires implementation |
| Operations inventory with sorting | Yes | No | Requires custom build |
| Setup time | Zero (built-in) | Hours | Days/weeks |

## Analytics use cases

### Incident response

#### Elevated error rate — find the source in minutes

**Scenario**  An on-call engineer receives an alert. They need to understand when the error spike started, which operations are affected, and which clients are impacted.

**How Cosmo handles it**  Open Metrics Analytics to see the error rate chart and identify when the spike began. Switch to Trace Analytics, group by error message, and click through to the individual traces with the highest error counts.

**Outcome**  Root cause identified in minutes rather than hours. The engineer can share timeline and scope with stakeholders before the incident escalates.

### Schema evolution

#### Deprecate a field without breaking clients

**Scenario**  A schema designer wants to remove an old field. They need to know which clients still use it and how frequently before making any change.

**How Cosmo handles it**  Open Schema Field Usage for the target field. Review which clients use it, how many requests they make, and when the field was last seen. Contact each client team with specific data about their usage before deprecating.

**Outcome**  The field is deprecated with full visibility into impact. Affected clients receive targeted communication — not a blanket announcement.

### Performance optimization

#### Find the slowest operations without guessing

**Scenario**  An engineering team has limited time for optimization and needs to know which operations are worth targeting.

**How Cosmo handles it**  Open Operations Tracking and sort by latency. Cross-reference with request count to find high-traffic, high-latency operations. Navigate directly to traces for any operation to investigate specific failures.

**Outcome**  Optimization effort is directed at the operations that affect the most users. Engineers stop guessing and start fixing.

### Client analysis

#### Isolate a client-specific issue without affecting others

**Scenario**  A mobile team reports slow responses. The platform team needs to confirm whether the issue is client-specific or systemic.

**How Cosmo handles it**  Use Client Identification headers to track the mobile client. Filter Metrics Analytics by client name and compare P95 latency and error rate against other clients. Check Operations Tracking for mobile-specific operations.

**Outcome**  The team confirms the issue is isolated to the mobile client and hands the investigation to the right team with specific data.

## Why teams use Cosmo Analytics

- **GraphQL context on every data point.** Operation name, operation type, client, and subgraph are first-class dimensions across every analytics view — not fields you have to add later.
- **Schema changes backed by real usage data.** Schema Field Usage shows exactly which clients and operations use every field, with first and last seen timestamps. Deprecation decisions stop being guesswork.
- **Connected views. No context switching.** Navigate from an operation directly to its traces. Filter by client and the selection carries through metrics, traces, and field usage. One investigation stays in one place.

## Full analytics for your federated graph