Your CDN cache hit ratio was sitting at 95%. Nobody touched anything, no alert fired, the site stayed up. Two weeks later the origin egress bill lands 40% over budget, and when you finally dig in, you find the ratio quietly slid to 72% right after a cache policy change nobody connected to cost. The site was technically healthy the whole time, so nothing alarmed. But every point the ratio dropped was more traffic hitting your origin, and the bill compounded a little more each day it went unnoticed.
That’s the silent drop, and it’s the single most expensive blind spot in edge operations, because cache hit ratio degrades in a way your monitoring is structurally built to ignore. This is a guide for operators who’ve lived that invoice: what actually causes cache hit ratio to fall, why standard monitoring never warns you, and how to treat the ratio as the production metric it is so you catch a drop in hours instead of on the statement.
What Is CDN Cache Hit Ratio?
Briefly, for shared vocabulary: cache hit ratio is the percentage of requests served straight from CDN edge cache without touching your origin. The formula is cache hits divided by total requests, times 100, so a 95% ratio means 95 of every 100 requests never reach your origin and 5 do. For static-heavy sites a healthy ratio lives in the 90 to 99% range, and a sustained figure below 80% usually points at a configuration problem rather than normal behavior. Most operators reading this know all that. The interesting part isn’t the definition; it’s what makes the number move and why nobody sees it move.
Why Cache Hit Ratio Is More Than a Performance Metric
Most writing about cache hit ratio treats it as a speed metric: higher ratio, faster pages, happier users. True, and incomplete. For anyone operating an edge, the ratio is simultaneously a CDN cost metric and a security signal, and those two framings are what justify watching it closely.
As a cost metric, it’s one of the most financially sensitive numbers in your stack, because every cache miss is a billable round trip to origin, paid twice as egress and as origin compute. A ratio drop doesn’t degrade cost gently; it scales your origin traffic against your full request volume, which is why a modest-looking percentage change produces an outsized bill. As a security and governance signal, the ratio is a tripwire. A sudden drop frequently means an unauthorized or unreviewed configuration change just landed, and a slow decline over weeks is often configuration drift in progress, settings that were tuned correctly at deployment quietly rotting as the application evolves and nobody re-checks the cache policy. Read that way, cache hit ratio deserves the same monitoring rigor you already give uptime and error rates, and almost no team gives it that.
What Causes CDN Cache Hit Ratio Drops?

Seven causes account for nearly every drop I’ve investigated. They divide cleanly into things your origin does, things your configuration does, and things your traffic does.
Cache-Control Header Misconfiguration
The most common cause by far, and the most insidious, because it originates upstream of your CDN settings. If your origin returns Cache-Control: no-cache, no-store, or private, the CDN honors it and declines to cache the response, no matter how aggressive your CDN-side rules are. A deployment that changes a framework default or flips a header on a route can kill caching for a whole path overnight, and because the change lives in your application’s response headers rather than your CDN dashboard, it’s the last place anyone thinks to look.
Short TTL Values
TTLs shorter than your content actually needs force the cache to expire and revalidate or refetch far more often than necessary. Content that changes weekly served with a TTL of minutes spends most of its life being re-fetched from origin for no reason. The waste is quiet because nothing breaks; the content is fresh, the site is fast, and the ratio is just lower than it should be, bleeding origin traffic the whole time.
Query String and Cookie Fragmentation
CDNs key the cache on the URL, so by default each unique URL is a separate cache entry. When your application appends timestamps, tracking parameters, or session tokens to URLs, the CDN sees a thousand variations of one asset as a thousand distinct objects, none of them warm. The cache fragments, the hit ratio falls, and the underlying content never changed at all. Marketing tracking parameters are a frequent and invisible culprit here.
Cache Purge Events
Purging forces refetch, so aggressive or frequent purging tanks the ratio during and after each event while the cache rewarms from origin. Teams that reflexively purge everything on every deployment or content update pay for it in a ratio that never gets to stay high, because they keep resetting it to cold. The purge felt free; the origin traffic that followed wasn’t.
Configuration Changes Nobody Tracked
Someone modifies a page rule, loosens a cache setting, or adjusts routing during an incident. The change reduces cache efficiency as a side effect nobody intended, it never gets documented, and the ratio drop that follows never gets connected back to it. This is the cause that turns a five-minute change into a month-long mystery, because the evidence and the cause live in different systems owned by different people.
Application Changes That Alter Traffic Patterns
A new feature, a redesigned page, or a modified endpoint can shift the composition of your traffic. Content that used to be static and cacheable now carries a dynamic element that bypasses cache, or a redesign moves users onto uncached paths. The ratio drops not because anything is misconfigured but because what you’re serving changed underneath a configuration tuned for the old shape.
Multi-CDN Cache Splitting
If you split traffic across providers, each maintains its own separate cache, so the same asset gets fetched from origin at least once per provider, and the long tail of less-popular content gets refetched far more because no single cache sees enough requests to stay warm. In a multi-CDN setup, more providers means more cold caches and a lower per-provider ratio. This one compounds with everything else and is covered in more depth in our guide on CDN egress costs, since it’s where cache splitting shows up on the bill.
Why Most Teams Miss Cache Hit Ratio Drops
Here’s the core of the problem, and it’s a detection gap, not a competence gap. Standard monitoring alerts on uptime, error rates, and response times, and a CDN cache hit ratio drop trips none of them. The site stays up, so uptime is fine. Error rates stay flat, because misses aren’t errors. Response times may tick up slightly as more requests go to origin, but rarely enough to breach a latency threshold. From the perspective of every alert you have configured, nothing happened.
The ratio is visible, technically. Your CDN dashboard shows it. But it’s a metric teams check after they already know something is wrong, not one that reaches out and tells them, so it functions as forensic evidence rather than an early warning. And even once you’re looking, there’s no native line connecting the drop to its cause: you can see the ratio fell and see that a cache policy was modified, but joining those two facts is manual investigation that, realistically, only happens once the bill forces it. That’s the whole trap, a financially critical metric that no alert watches and no system explains. This is exactly the metric drift Lelops Sentry is built to catch, monitoring your CDN cache hit ratio continuously and firing an alert the moment it drops below your baseline, with the configuration change that caused it attached, so you can revert before the cost compounds.
How to Detect and Prevent Silent Cache Hit Ratio Drops

The fix is to stop treating the ratio as something you check and start treating it as something that watches you back. Six practices, roughly in order of impact.
Monitor CDN Cache Hit Ratio as a First-Class Production Metric
Give it the urgency you give uptime. Establish a baseline for normal, and alert when the ratio falls below it by any meaningful margin. The entire failure mode above depends on nobody being told; an alert closes it. Waiting for the monthly egress bill to surface the drop is choosing to find out in the most expensive way possible.
Correlate Ratio Changes with Configuration State
When the ratio drops, the first and best question is “what changed in the last 24 hours?” If you can see the configuration change that immediately preceded the drop, diagnosis collapses from days of investigation to a minute of reading. The value isn’t the alert alone; it’s the alert arriving with the cause already attached, so you’re deciding whether to roll back rather than starting to investigate.
Set Alerts on Rate of Change, Not Just Absolute Thresholds
A ratio steady at 88% might be perfectly normal for your application. A ratio that fell from 95% to 88% in twelve hours is an incident. Absolute thresholds miss the second case entirely, because 88% clears the bar. Alert on the delta and the velocity, not just the floor, so a fast slide from a healthy number triggers attention even while the number still looks acceptable.
Audit Cache-Control Headers After Every Deployment
Since header changes are the top cause and the hardest to spot, build a check into your deployment pipeline that verifies the response headers on key assets haven’t changed. A single flipped Cache-Control directive can invalidate your whole cache strategy silently, and catching it at deploy time turns a two-week mystery into a failed pipeline step.
Review Cache Purge Frequency and Scope
If your team purges often, ask whether full purges are actually necessary or whether targeted, path-scoped purges would deliver the freshness you need without dumping the entire cache. Most reflexive full purges are broader than the content that actually changed, and narrowing them preserves the ratio while still shipping updates.
Normalize Query Strings and Cache Keys
Configure the CDN to ignore the query parameters that don’t affect the response, tracking IDs, timestamps, session tokens, when computing the cache key. This collapses the thousand fragmented variations of one asset back into a single warm cache entry, and it’s often the single highest-leverage fix for a ratio that’s mysteriously low despite correct TTLs.
CDN Cache Hit Ratio Benchmarks by Content Type
Benchmarks are only meaningful per content type, because a single blended ratio hides the story. Use these as rough targets, not universal law, and calibrate against your own traffic:
| Content type | Typical healthy ratio |
|---|---|
| Static assets (images, CSS, JS, fonts) | 95–99% |
| HTML pages (mostly static) | 85–95% |
| Cacheable API responses | 60–85%, depending on TTL strategy |
| Dynamic or personalized content | Not expected to cache; exclude from the metric |
The last row matters more than it looks. Dynamic, per-user content was never meant to be cached, and leaving it in your ratio calculation drags the number down and hides the health of the content that should be cached. Measure the ratio on cacheable traffic, and for a static-heavy site, treat a sustained figure for your CDN cache hit ratio below 80% as a configuration issue to investigate rather than a baseline to accept.
Lelops Sentry monitors your cache hit ratio continuously and alerts you the moment it drifts below your baseline, so you catch the drop before the egress bill does: lelops.ai/product/sentry.
FAQ
What is a CDN cache hit ratio?
A CDN cache hit ratio is the percentage of requests served directly from the CDN’s edge cache without reaching your origin server, calculated as cache hits divided by total requests times 100. A ratio of 95% means 95 of every 100 requests are answered at the edge and only 5 travel to origin. It is both a performance indicator and a cost indicator, since every request that misses the cache becomes a billable trip to your origin.
What is a good cache hit ratio for a CDN?
A good ratio depends on content type: static assets like images and scripts should sit around 95 to 99%, mostly-static HTML around 85 to 95%, and cacheable API responses roughly 60 to 85% depending on TTL strategy. Truly dynamic or personalized content is not expected to cache and should be excluded from the calculation. For a static-heavy site, a sustained ratio below 80% generally signals a configuration problem worth investigating.
What causes a sudden drop in cache hit ratio?
Sudden drops are most often caused by a Cache-Control header change from the origin that tells the CDN not to cache, a cache policy or page rule modification, or a large cache purge that forces refetching from origin. Query string fragmentation from newly added tracking or session parameters can also cause a sharp drop by turning one cacheable asset into many uncached variations. Because these changes often accompany a deployment or an incident fix, the drop frequently traces back to a configuration change made shortly before it.
How do I monitor my CDN cache hit ratio in real time?
Real-time monitoring means treating the ratio as a first-class production metric with a defined baseline and automated alerts, rather than a number you check after a problem surfaces. Effective monitoring alerts on the rate of change, not just an absolute floor, so a fast slide from a healthy number is caught even while the value still looks acceptable. Correlating the ratio with recent configuration changes is what turns an alert into an immediate diagnosis instead of a fresh investigation.
Does a low cache hit ratio increase my origin egress costs?
Yes, directly and often steeply. Every request that misses the cache is served from your origin and billed as egress and origin compute, so a lower ratio means more origin traffic on the same user volume. Because the extra traffic scales against your total requests, even a modest ratio drop can produce a disproportionate increase in monthly egress spend, which is why an unnoticed drop so often surfaces first as an unexpected bill.
Christophe Pronier


