CDN analytics turns delivery data into operational decisions. This guide explains which CDN performance metrics to monitor, how to connect changes to likely causes, and how to use a repeatable monthly review to improve cache behavior, origin response time, reliability, and user experience.
Overview
A CDN dashboard can expose dozens of charts, but not every metric deserves equal attention. The useful ones answer a practical question: are requests being served efficiently from the edge, and when they are not, why? A good monitoring routine combines delivery, origin, reliability, geography, and security data rather than relying on a single headline number.
Start by recording a baseline for a representative period. Depending on your traffic volume, this might be the previous 7, 14, or 30 days. Note the date range, CDN configuration, deployment activity, major campaigns, and unusual traffic events. Without that context, a change in request volume or cache hit ratio can be mistaken for a configuration problem.
Provider terminology varies. One platform may label a response as a cache hit, while another reports cache status through request logs or separates edge hits from shield hits. Before building alerts or reports, confirm what each metric includes, which traffic is excluded, and whether values are sampled.
What to track
The following metrics form a practical minimum for website delivery monitoring. Track both the absolute value and the trend compared with a previous period.
| Metric | What it shows | What to investigate when it changes |
|---|---|---|
| Cache status and cache hit ratio | How often eligible requests are served from an edge cache instead of the origin. | Unexpected bypass rules, short freshness periods, cookies, query strings, authorization headers, purges, or assets that should not be cached. |
| Bandwidth and egress | How much data the CDN delivers and which paths or file types consume it. | Large media files, unoptimized images, duplicate downloads, traffic growth, or a deployment that changed asset behavior. |
| Request volume | The number and distribution of requests received by the edge. | Traffic campaigns, bots, retry loops, missing asset caching, API polling, or a sudden change in user behavior. |
| Origin response time | How long the origin takes to respond when the CDN contacts it. | Slow database work, overloaded hosting, uncached HTML, upstream dependencies, connection limits, or a recent application release. |
| TTFB and edge response time | How quickly users receive the first response bytes from the delivery path. | Origin latency, poor routing, cache misses, geographic distance, TLS or connection issues, or an overloaded edge location. |
| HTTP error rates | The proportion and type of failed requests, such as 4xx and 5xx responses. | Broken deployments, origin outages, rate limits, invalid cache rules, authentication failures, or upstream timeouts. |
| Geographic performance | Whether delivery quality differs by country, region, provider, or point of presence. | Routing problems, regional origin distance, local connectivity, DNS configuration, or a location-specific edge incident. |
| Security events | Requests challenged, blocked, rate-limited, or identified as suspicious. | Attack traffic, an overly broad rule, false positives, credential abuse, or a change in application exposure. |
Cache metrics need context. A low overall hit ratio is not automatically a failure if much of the traffic consists of personalized pages, checkout requests, uncached APIs, or content that changes frequently. Segment the data by hostname, URL path, status code, file type, and cache status before changing policy. Static CSS, JavaScript, fonts, and versioned images usually deserve a different analysis from HTML, account pages, or transactional endpoints.
Origin response time is particularly useful when paired with cache status. If cache misses are slow but hits remain fast, the edge is working and the origin path needs attention. If both hits and misses are slow, investigate network routing, TLS negotiation, edge configuration, or the measurement method. For a broader investigation of first-byte delays, see the guide to improving TTFB.
Keep a small set of dimensions in every report: hostname, path group, country or region, device category where available, response status, and cache outcome. A single site-wide average can hide a failing checkout route or a slow regional segment.
Cadence and checkpoints
Use different review intervals for different types of risk. Real-time or near-real-time alerts are appropriate for availability and severe error changes. A daily check can catch incidents and traffic anomalies. A monthly review is better for identifying gradual cache degradation, rising origin load, or inefficient asset delivery.
After a deployment or configuration change
Compare the period immediately before and after the change. Check cache status, purge activity, origin response time, 4xx and 5xx rates, and the most requested paths. Confirm that cache-control headers, cookies, query-string handling, and asset versioning behave as intended. If you changed cache rules, test representative URLs from more than one location rather than relying only on the dashboard.
During a traffic spike
Watch request volume, bandwidth, cache hit ratio, origin connections, origin response time, and error rates together. A traffic increase with stable origin load may indicate effective edge caching. A similar increase in origin requests suggests that content is bypassing the cache or being repeatedly purged. For availability planning, pair analytics with a documented CDN failover testing process.
Monthly review
Record the current period, comparison period, traffic changes, deployments, incidents, and three actions. A concise reporting template can look like this:
- Delivery: cache ratio, request volume, bandwidth, and top paths.
- Origin: median or percentile response time, origin traffic, and slowest route groups.
- Reliability: 4xx and 5xx trends, timeout events, and affected regions.
- Security: blocked or challenged requests and notable rule changes.
- Actions: one configuration test, one application or asset optimization, and one item to monitor next month.
How to interpret changes
Investigate relationships, not isolated movements. A falling cache hit ratio combined with rising origin requests is more informative than either metric alone. It may indicate a cache-control change, a new cookie, a query-string variation, or an increase in uncached content. Review response headers and request logs for affected URL groups before purging the entire site.
A rise in bandwidth can come from more visitors, larger files, repeated downloads, or a change in compression. Compare bandwidth with request volume and average response size. If bandwidth rises while requests remain stable, inspect large responses and recently changed media. If both rise, first determine whether the change is expected traffic or an efficiency problem.
When error rates increase, separate client and server errors. A rise in 4xx responses may reflect broken links, rejected authentication, a bot pattern, or an application change. A rise in 5xx responses and origin timeouts usually warrants faster investigation of the origin and its dependencies. Segment errors by URL, region, and status code to avoid treating unrelated issues as one incident.
Geographic differences require careful testing. A slower region may reflect the visitor's access network, a routing choice, origin distance, or a local edge issue. Validate the pattern with repeated tests and compare it with real-user data when available. Avoid changing DNS or CDN routing based on one synthetic test.
For cache investigations, use response headers, browser developer tools, and controlled requests alongside dashboard data. The website caching test guide explains how to distinguish browser, CDN, and origin behavior. If the hit ratio is persistently low, use the cache hit ratio troubleshooting guide to inspect eligibility, bypasses, and cache keys.
When to revisit
Revisit this monitoring plan at least monthly and after any event that changes traffic or delivery behavior. That includes a CDN migration, DNS change, hosting move, major application release, new authentication flow, image or video rollout, cache-rule update, security policy change, or sustained traffic growth.
Refresh the dashboard definitions twice a year or whenever your provider changes its analytics interface. Confirm whether labels, sampling, aggregation windows, cache-status categories, and geographic reporting have changed. Keep a record of metric definitions so that a year-over-year comparison remains meaningful.
Use the next review to turn observations into one controlled test. Examples include extending the freshness period for versioned static assets, separating personalized HTML from public HTML, reducing unnecessary API polling, improving image delivery, or adjusting an origin capacity limit. Document the hypothesis, affected paths, expected result, test window, and rollback method.
For practical follow-up, review your report in this order:
- Check availability and 5xx errors first.
- Compare cache status with origin traffic and origin response time.
- Identify the paths, regions, or file types driving the change.
- Verify headers and cache behavior with a direct test.
- Choose one low-risk optimization and schedule a comparison.
- Record the result before making another change.
This routine keeps CDN analytics tied to decisions rather than passive reporting. For implementation details, use the cache-control headers guide, the cache-busting guide, or the Cloudflare caching rules guide when the data points to a policy or cache-key problem.