Choosing a CDN for a small or medium-sized business is less about finding a universal winner and more about comparing delivery cost, cache behavior, integration effort, and measurable origin performance. This guide provides a repeatable way to compare managed caching solutions, estimate a 12-month cost, test edge latency and TTFB, and migrate with a rollback plan.
Overview
A CDN or reverse proxy cache can serve static assets and, where your application permits it, cached HTML or API responses from an edge location closer to visitors. The result may be a faster first response, lower origin bandwidth, and better resilience during traffic spikes. Those benefits depend on cache-control headers, cache rules, object size, visitor geography, and the percentage of requests that can safely be cached.
For an SMB website, the best CDN is usually the one that meets four practical requirements: predictable total cost, acceptable performance in the markets you serve, a safe integration path, and support that matches your operational needs. A low headline rate is not necessarily the lowest annual cost if requests miss the cache, purge operations are limited, or the platform requires substantial engineering work.
Provider comparison framework
The table below is deliberately framed as a buying worksheet rather than a ranking. Pricing, service levels, network coverage, and product features change. Confirm each item on the provider's current pricing and contract pages before making a purchase.
| Provider | Useful comparison angle | Origin pull | Cache-hit expectation | Pricing and egress questions | Support or SLA check |
|---|---|---|---|---|---|
| Cloudflare | Managed DNS, edge rules, security, and broad self-service tooling | Confirm origin type, request handling, and cache rules for HTML and APIs | Measure by URL class; do not assume the site-wide ratio | Check plan limits, included services, traffic treatment, and overage terms | Compare the support level and any SLA included with the selected plan |
| Fastly | Fine-grained cache control, purge workflows, and developer-oriented configuration | Test shielding, revalidation, headers, and origin fail behavior | Measure after warming and during normal traffic | Request a written estimate for bandwidth, requests, purge use, and add-ons | Review contractual response targets and escalation procedures |
| Bunny | Simple CDN deployment and usage-based comparison for many asset-heavy sites | Confirm origin pull settings, storage options, and fallback behavior | Separate cacheable media from dynamic page requests | Check regional delivery rates, minimums, storage, and purge-related charges | Verify available support channels and any formal service commitment |
| KeyCDN | Straightforward pull-zone and cache configuration comparison | Test pull-zone headers, query strings, and cache expiration | Use response headers and logs to establish a baseline | Check traffic pricing by region, request costs, and add-on features | Confirm support scope, response expectations, and escalation options |
| Amazon CloudFront | Integration with AWS origins, controls, and adjacent cloud services | Test origin groups, behaviors, signed requests, and invalidations | Calculate separately for each behavior or path pattern | Model data transfer, requests, invalidations, and related AWS services | Compare the support plan and any availability commitments in writing |
| Google Cloud CDN | Integration with Google Cloud load balancing and application infrastructure | Test backend configuration, cache modes, and revalidation | Measure by backend and cache mode | Model delivery, requests, cache fill, and associated Google Cloud costs | Check the selected support package and applicable service commitments |
| Azure Front Door | Azure integration, routing, edge delivery, and security controls | Test routes, rules, origins, health probes, and cache behavior | Compare static, HTML, and API paths independently | Review bandwidth, requests, rules, security features, and regional charges | Confirm support plan and the service terms attached to the deployment |
| Akamai | Enterprise delivery, security, and contract-based managed operations | Discuss origin architecture, shielding, failover, and cache purging | Agree on measurement definitions and reporting before launch | Request a complete quote covering traffic, services, support, and commitments | Review negotiated SLA, support coverage, and incident process |
“Typical TTFB” and “typical edge latency” are not portable provider scores. They vary with the visitor's location, DNS path, TLS negotiation, object size, origin distance, and whether the request is a cache hit. Establish your own baseline with the same URLs, regions, protocol, and test conditions. For measurement guidance, see the CDN analytics guide and the guide to improving TTFB.
How to estimate
Use a 12-month model instead of comparing a single advertised rate. Start with these formulas:
- Annual delivered volume: monthly CDN traffic × 12.
- Annual CDN cost: delivery charges + request charges + storage or add-ons + support or plan fees + estimated overages.
- Origin bandwidth: total cacheable traffic × (1 − cache-hit ratio), plus traffic that is intentionally bypassed.
- Origin reduction: baseline origin requests or bytes − post-CDN origin requests or bytes.
- Effective cost per delivered GB: annual CDN cost ÷ annual delivered GB.
Use separate rows for static assets, HTML, media, downloads, and APIs. A single blended cache-hit ratio can conceal a problem: a site may cache nearly all images while sending every HTML request to the origin. Record hits, misses, bypasses, purges, and stale responses where the provider exposes them. The related cache-hit ratio guide explains how to investigate a low result.
Decision flowchart
Choose by budget: need a simple, low-commitment deployment? Compare usage-based totals, included features, minimums, and support costs using a representative traffic sample.
Choose by latency: are visitors spread across regions or is TTFB the main issue? Run identical tests from your priority markets, then compare cache-hit TTFB, miss TTFB, DNS time, and origin response time.
Choose by integration: does the site depend on AWS, Azure, Google Cloud, WordPress, WooCommerce, private APIs, or custom deployment pipelines? Shortlist providers that fit your existing DNS, certificate, purge, logging, and access-control workflows.
Inputs and assumptions
Before requesting quotes, assemble twelve months of traffic if available. Capture monthly transferred bytes, request count, top URLs, visitor countries, peak requests per second, origin bandwidth, current TTFB, and cache-control headers. Also record how often content is published, how quickly a purge must take effect, and which paths must never be cached.
For WordPress, identify logged-in sessions, preview URLs, search pages, cookies, admin paths, and WooCommerce cart, checkout, account, and payment flows. A CDN for WordPress performance should normally cache versioned CSS, JavaScript, fonts, and images first. Add public HTML only after testing cookies, personalization, forms, and invalidation. The WordPress caching guide provides a path-by-path model.
Assume cacheability conservatively. If you do not have reliable data, create low, expected, and high cases rather than selecting one optimistic percentage. Include operational costs: engineering time for DNS and rule changes, monitoring, purge automation, incident response, and the cost of keeping the old origin configuration available during the trial.
Worked examples
Consider an illustrative SMB site that delivers 600 GB per month and currently sends 100% of page requests to its origin. Suppose a test indicates that 85% of transferred bytes are cacheable and that the CDN produces a 90% hit rate on those cacheable bytes. The estimated origin reduction for that portion is:
600 GB × 85% × 90% = 459 GB per month avoided at the origin.
This is a planning example, not a provider benchmark. It does not mean origin traffic falls by 90% overall: uncached HTML, APIs, cookies, misses, revalidations, and bypassed paths remain. Add the site's non-cacheable traffic to the estimate before sizing the origin.
For cost modeling, create three scenarios. In a low case, use the lower traffic forecast and conservative request growth. In the expected case, use the median of the last several comparable months. In a high case, model a campaign or seasonal peak and include a safety margin. Then place each provider's verified rates and plan fees into the same spreadsheet. If a provider quotes rather than publishes a rate, label the value as a quote and record the quote date, included volume, regions, exclusions, and renewal terms.
Run a controlled pilot before committing. Put a test hostname behind the candidate CDN, fetch the same assets from representative locations, confirm cache-status headers, inspect origin logs, test purges, and compare warm-cache and cold-cache TTFB. Use the cached-site testing guide to distinguish a browser cache from an edge cache.
When to recalculate
Recalculate this comparison at least every six months and immediately after a provider changes pricing, traffic definitions, cache-purge behavior, network coverage, or product packaging. Repeat it before an annual renewal, a major campaign, a hosting move, or a redesign that changes asset size and URL structure.
During each review, export the last period's delivered bytes, requests, cache-hit ratio, purge count, origin bandwidth, p50 and p95 TTFB, error rate, and support incidents. Compare those results with the assumptions in your original model. Investigate any gap before switching providers; a low hit ratio may result from query strings, cookies, short TTLs, missing cache-control headers, or an incorrect bypass rule rather than the CDN itself.
Migration and rollback checklist
- Inventory DNS records, certificates, origins, redirects, cache rules, protected paths, and purge integrations.
- Document a baseline for TTFB, availability, origin load, cache-hit ratio, and key user journeys.
- Configure the candidate CDN on a test hostname. Start with static assets and explicitly bypass logged-in, transactional, and sensitive paths.
- Validate headers, compression, range requests, redirects, CORS, cookies, image delivery, forms, APIs, and WooCommerce transactions if applicable.
- Warm representative URLs, test purges and revalidation, and compare origin logs with CDN analytics.
- Lower DNS TTL in advance of the cutover, while recognizing that resolvers may retain records beyond the requested TTL.
- Change DNS or routing during a monitored window. Keep the previous origin and configuration available.
- Watch errors, TTFB, cache misses, origin saturation, checkout, login, webhooks, and support reports for at least one normal traffic cycle.
- If a critical regression appears, restore the previous DNS or routing target, purge incorrect objects, and document the failure before retrying.
- After stabilization, raise appropriate TTLs, automate cache purges, and schedule the next cost and performance review.
For higher availability requirements, pair the migration plan with a documented failover test; the CDN failover guide covers that process. Keep this worksheet with your renewal records so the next comparison uses measured traffic and outcomes rather than assumptions.