The CDN math that actually matters for e-commerce
Forget the marketing slides. Here are the four numbers that determine whether your CDN is paying for itself.
When we audit a CDN deployment for an e-commerce client, we don’t start with the dashboard the vendor provides. We start with four numbers. If these four numbers aren’t being tracked, optimized, and reported every month, the deployment is almost certainly underperforming.
1. Cache hit ratio (by content type)
The aggregate cache hit ratio is mostly useless — it’s dominated by static assets that would cache anywhere. What matters is the cache hit ratio broken down by content type:
- Static assets (JS, CSS, fonts, images): should be 99%+
- Catalog HTML: should be 70-90%, depending on personalization
- Product detail pages: typically 60-85%, with proper Vary handling
- API responses: highly variable, but anywhere from 20-60% is achievable for read-heavy endpoints
If your catalog hit ratio is below 70%, there’s almost certainly headroom — usually in cookie handling, Vary headers, or cache key construction.
2. P75 Time to First Byte (by region)
P50 latency hides regional problems. We focus on P75 TTFB segmented by region. If your customers in Singapore see P75 TTFB of 800ms while your London users see 120ms, your routing is broken — and the conversion data will agree.
3. Origin egress cost trend
The single most direct measure of CDN value. If your origin egress is flat or growing while your traffic grows, your CDN isn’t doing its job. We expect to see origin egress fall relative to total traffic as a deployment matures — sometimes by 40-70%.
4. Core Web Vitals (real-user, not synthetic)
Synthetic monitoring is useful but lies. Real-user CWV from the field is the only number that correlates with revenue. We track LCP, INP, and CLS at P75 for the segments that matter (logged-in users, mobile, key geos).
Why these matter together
Any one of these can look fine in isolation. They have to be optimized together — pushing cache hit ratio up by ignoring cache freshness destroys conversion. Lowering TTFB with aggressive prefetching can wreck CLS. The art is balancing all four.
If you’d like a vendor-neutral look at where your CDN deployment sits on these four dimensions, book a consultation. We typically find 2-3 specific optimization opportunities in the first 30 minutes.