CDN bandwidth is often your largest infrastructure line item — and the most controllable. Unlike compute, you pay for every byte delivered. With the right caching, compression, and delivery strategy, most organizations can reduce CDN bandwidth costs by 30–60% without degrading performance.
CDN pricing looks simple: dollars per terabyte. In practice, your bill is driven by cache efficiency, origin behavior, content weight, and traffic quality. A site delivering 50 TB/month at $0.08/GB pays $4,000 monthly. Cut billable egress by 40% and you save $19,200 annually — budget you can reinvest in edge performance. Below are 12 field-tested techniques we use to optimize CDN spend for high-traffic sites.
How CDN Billing Actually Works
Most providers charge for three things: edge bandwidth (GB delivered to users), HTTP requests (per 10k), and origin shield/egress. Bandwidth dominates, typically 70–85% of the invoice. You are billed for the compressed bytes served from edge, plus any origin fetches when content is not cached. Low cache hit ratio, large unoptimized assets, and bot traffic inflate all three simultaneously.
Regional pricing also matters. North America and Europe are cheapest ($0.02–$0.08/GB at scale), while South America, Australia, and parts of Asia can be 2–3x higher. If 20% of your audience is in high-cost regions, optimizing delivery there yields disproportionate savings. Understanding this breakdown helps you prioritize fixes with the highest ROI.
12 Proven Techniques to Reduce CDN Bandwidth Costs
1. Maximize Cache Hit Ratio with Tiered Caching
Cache hit ratio (CHR) is the single biggest lever. At 80% CHR, 20% of requests hit origin; at 95% CHR, only 5% do — a 75% reduction in origin egress. Enable tiered or hierarchical caching so a miss at an edge POP checks a regional shield POP before contacting origin. This collapses duplicate origin requests globally. Also cache 404s and redirects briefly (30–60s) to avoid origin floods from scanners.
| Metric | Edge Processing | Origin Processing | Improvement |
|---|---|---|---|
| Latency (p50) | 8–15 ms | 120–300 ms | 15–20x faster |
| Cache Hit Ratio | 85–95% | N/A | Reduces origin load |
| TLS Handshake | 1–5 ms (nearby PoP) | 50–150 ms | 10–30x faster |
| Bandwidth Cost | $0.01–0.04/GB | $0.05–0.12/GB | 3–5x cheaper |
| Availability | 99.99% (multi-PoP) | 99.9% (single DC) | 10x fewer outages |
2. Deploy Origin Shield Strategically
Origin Shield designates one POP as the sole gateway to your origin. Without it, 50 edge POPs might each fetch the same object. With it, only one fetch occurs and the other 49 pull from shield cache. Most providers charge slightly for shield requests but save far more on origin egress and compute. Place your shield close to your origin region and enable shield-to-shield failover for resilience.
3. Enable Brotli and Zstandard Compression
Gzip is baseline, but Brotli (level 4–6) delivers 15–25% smaller text files with similar CPU cost, and Zstandard excels for dynamic JSON/API responses. Enable Brotli at the edge and ensure your origin serves pre-compressed assets when possible. For APIs, negotiate Accept-Encoding: br, zstd and compress on the fly. This is especially effective for JavaScript, CSS, and JSON where savings are highest.
load_module modules/ngx_http_brotli_filter_module.so;
brotli on;
brotli_comp_level 5;
brotli_types text/css application/javascript application/json image/svg+xml;
# Cache compressed variants at CDN edge
add_header Vary Accept-Encoding;
add_header Cache-Control "public, max-age=31536000, immutable";
4. Optimize Images and Video Aggressively
Media dominates bandwidth. For ecommerce and media sites, unoptimized images can be 70% of egress. Convert to AVIF (30–50% smaller than JPEG) with WebP fallback, implement responsive images with srcset and client hints, and lazy-load below-the-fold. For video, use adaptive bitrate (HLS/DASH) with per-title encoding and cap maximum bitrate for mobile. Edge image optimization services can automate format negotiation and resizing, cutting media bandwidth by 40–60%.
5. Tune Cache-Control and TTLs Precisely
Default TTLs are often too short. Use long TTLs for immutable assets (hashed filenames: 1 year), medium TTLs for semi-static content (product images: 1 day to 1 week), and short but non-zero TTLs for dynamic pages (60–300s with stale-while-revalidate). Set Cache-Control: public, max-age=86400, stale-while-revalidate=300, stale-if-error=86400 to serve stale content during revalidation, avoiding origin thundering herds. Version URLs instead of purging when possible.
# Immutable assets (app.abc123.js)
Cache-Control: public, max-age=31536000, immutable
# Product API - short cache with background revalidation
Cache-Control: public, max-age=60, stale-while-revalidate=300
CDN-Cache-Control: max-age=300 # CDN caches longer than browser
6. Minify, Bundle, and Tree-Shake Assets
Every kilobyte saved is multiplied by millions of requests. Minify CSS/JS/HTML, remove unused CSS with PurgeCSS, and tree-shake JavaScript bundles. Enable HTTP/2 or HTTP/3 multiplexing so bundling strategy can be less aggressive — serve smaller, cache-friendly chunks rather than one monolithic bundle that invalidates frequently. Pre-compress with brotli -q 11 for static assets at build time.
7. Control Prefetch and Preload Carefully
Aggressive prefetching (e.g., <link rel="prefetch"> for every link) wastes bandwidth on pages users never visit. Audit prefetch hit rates; if <30% of prefetched resources are used, disable it. Use speculative prefetch only for high-probability next navigations (e.g., checkout step 2 after step 1). Similarly, limit preload to critical fonts and hero images. Prefetching should improve LCP, not inflate your bill.
8. Leverage Edge Computing for Request Collapsing
Use edge workers to collapse duplicate origin requests, normalize query strings, and serve stale content intelligently. For example, strip marketing UTM parameters from cache keys so /product?id=123&utm_source=email and /product?id=123 share one cache entry. Edge functions can also generate lightweight responses (like resized images or API aggregation) without an origin round trip, eliminating egress entirely for those requests.
9. Filter Bots, Scrapers, and Bad Traffic
Bad bots can account for 15–40% of CDN requests. They inflate bandwidth and request charges while never converting. Implement bot management, rate limiting, and a properly tuned WAF to challenge or block scrapers. See our CDN security guide and DDoS protection strategies for filtering without blocking legitimate crawlers. Also block hotlinking and enforce signed URLs for premium assets to prevent bandwidth theft.
10. Negotiate Pricing and Consider Multi-CDN
Published pay-as-you-go rates are 2–5x higher than committed-use discounts. If you exceed 10 TB/month, request custom pricing, commit discounts, or burst allowances. For global audiences, a multi-CDN with intelligent load balancing lets you route traffic to the cheapest performant provider per region. Use real user monitoring (RUM) to shift traffic dynamically — e.g., serve Europe via Provider A at $0.03/GB and South America via Provider B at $0.06/GB.
11. Purge Selectively, Not Globally
Global purges (purge all) drop CHR to zero and cause an expensive origin stampede. Use surrogate keys or cache tags to purge only affected objects. For example, tag all product pages with product-123 and purge that tag on update, leaving the rest of the cache intact. Soft purges (serve stale while revalidating) are cheaper than hard purges. Automate purges via CI/CD webhooks rather than manual clears.
12. Monitor, Attribute, and Right-Size Continuously
You cannot optimize what you do not measure. Break down bandwidth by content type, URL pattern, country, and cache status (HIT vs MISS) daily. Set alerts for CHR drops below 90%, origin egress spikes, or bot ratio increases. Use log delivery and analytics to identify top 20 URLs driving 80% of cost — often a few uncacheable API endpoints or oversized hero videos. Review reserved capacity monthly and downgrade over-provisioned tiers.
| Technique | Typical Savings | Effort | Best For |
|---|---|---|---|
| Tiered Cache + Origin Shield | 25–40% origin egress | Low | All sites, immediate win |
| Brotli / Zstd Compression | 12–20% bandwidth | Low | Text-heavy, JS/CSS/JSON |
| AVIF/WebP + Responsive Images | 40–60% media bandwidth | Medium | Ecommerce, media, blogs |
| TTL Tuning + SWR | 15–30% requests | Low | Dynamic + semi-static |
| Bot Filtering & WAF | 10–30% total requests | Medium | Sites with high bot traffic |
| Multi-CDN Negotiation | 20–30% blended $/GB | High | 50TB+/mo, global audience |
Putting It Together: A 30-Day Cost Reduction Plan
Week 1: Audit CHR, top bandwidth URLs, and bot ratio. Enable Origin Shield, tiered caching, and Brotli. Fix obvious TTL misconfigurations.
Week 2: Deploy image optimization (AVIF/WebP, srcset, lazy-load) and minify assets. For ecommerce, prioritize product images and category pages — see CDN for ecommerce optimization for implementation patterns.
Week 3: Implement cache-key normalization, surrogate-key purging, and rate limiting. Review prefetch/preload usage and disable wasteful hints.
Week 4: Analyze regional cost breakdown, negotiate committed pricing, and evaluate multi-CDN if spend exceeds $5k/month. Set up ongoing dashboards for CHR, egress, and cost per 1k visits.
Most teams see 20–30% savings in the first two weeks from caching and compression alone. Media optimization and bot filtering deliver the next 15–25% over the following month.
Frequently Asked Questions
What is the biggest driver of high CDN bandwidth bills?
Low cache hit ratio is the primary driver. Every cache miss forces the CDN to fetch from origin, incurring both origin egress and CDN bandwidth charges. Raising hit ratio from 85% to 95% can cut bandwidth costs by 30–40% alone, making tiered caching and Origin Shield the highest-ROI fixes.
Does enabling Brotli or Zstandard really reduce CDN costs?
Yes. Brotli compresses text assets 15–25% better than Gzip, and Zstandard is even faster for dynamic content like JSON APIs. For text-heavy sites, modern compression typically reduces billable bandwidth by 12–20% with no visual quality loss and minimal CPU overhead at the edge.
How does image optimization impact CDN bandwidth?
Images and video often account for 60–75% of bandwidth. Converting to AVIF/WebP, implementing responsive sizing with srcset, and using adaptive bitrate video can cut media bandwidth by 40–60%. This is the largest saving opportunity for ecommerce, travel, and media sites.
Should I use a multi-CDN to reduce costs?
A multi-CDN with intelligent load balancing lets you route traffic to the cheapest performant provider per region and negotiate volume discounts. It adds operational complexity but can save 20–30% for sites spending over $5k/month while improving resilience and reducing vendor lock-in.
Reducing CDN bandwidth cost is not about serving less — it is about serving smarter. Cache more, compress better, deliver only the bytes users need, and pay the right price per gigabyte. Start with hit ratio and compression, then layer on media and traffic quality optimizations. Your users get faster pages and your finance team gets a smaller invoice.