The First Thing I Wish Someone Told Me About Cloudflare Bandwidth Savings
When a client first asked me how to calculate Cloudflare bandwidth savings, I did what most engineers do: I opened the Cloudflare dashboard and hunted for a bandwidth line item on the invoice. It wasn’t there. That missing line is the entire point—Cloudflare does not charge per gigabyte for CDN delivery on its standard plans.
This directly answers the persistent query: Does Cloudflare charge for bandwidth? For the proxy and CDN service, no. You pay a flat plan fee (Free, Pro, Business, or Enterprise) and receive unmetered DDoS protection and caching. The dollars you save are the ones you no longer pay your origin cloud provider for data transfer out (egress).
I learned this the hard way in 2022 while migrating a video‑heavy news site doing roughly 40 TB/month. We expected a Cloudflare credit; instead the AWS bill dropped from $3,600 to $1,100 in month one. The most people don’t realize insight: Cloudflare’s value isn’t “free bandwidth”—it’s origin egress avoidance.
Another common question: Is there a bandwidth limit for Cloudflare CDN? On Free, Pro, and Business plans, Cloudflare advertises unmetered mitigation and no hard caps, but they reserve fair‑use rights for abusive traffic spikes. Enterprise contracts often commit specific bandwidth. For 99% of SMBs, you won’t hit a wall, but you’re not entitled to infinite saturation either.
And to satisfy scale curiosity: What percentage of Internet goes through Cloudflare? Cloudflare’s own Radar data indicates they handle roughly 20% of global web requests, though that varies by region and service. See the Cloudflare Radar 2022 review for methodology and caveats.
The thing nobody tells you about: that 20% figure is request‑based, not bandwidth‑based. Video‑heavy CDNs like Netflix inflate bandwidth share elsewhere, so Cloudflare’s bandwidth share is lower. Don’t use the percentage to size your own savings; use your own logs.
The Origin Egress Savings Worksheet: A Reusable Framework
To move beyond hand‑waving, I built a reusable spreadsheet I call the Origin Egress Savings Worksheet. It forces you to input three real numbers: current monthly origin egress in GB, cloud provider’s per‑GB egress rate, and Cloudflare Cache Hit Ratio (CHR) weighted by bytes.
Here is the framework as a table you can copy:
| Input | Source | Example |
|---|---|---|
| Total bandwidth served to users (GB) | Cloudflare Analytics → Traffic | 10,000 |
| Cache Hit Ratio by bytes (%) | Cache Analytics → Bandwidth view | 85% |
| Origin egress rate ($/GB) | AWS/GCP bill | 0.09 |
| Bandwidth Alliance discount factor | Partner config | 0% or 100% |
The core formula is simple: Saved Origin GB = Total Served GB × CHR. Then Monthly Savings = Saved Origin GB × Origin $/GB × (1 − BA discount). In the example, 10,000 × 0.85 = 8,500 GB offloaded, worth $765/month before any partner waiver.
If you’d rather not maintain Excel, we built the Cloudflare Bandwidth Savings Calculator that mirrors this worksheet and adds Bandwidth Alliance logic automatically.
The trap I fell into: dashboard CHR is often request‑weighted. A hit on a 2 KB CSS file counts the same as a miss on a 50 MB video. You must export the “Bandwidth” panel to weight by bytes, or your projection will be 30–50% too optimistic.
For a deeper pre‑migration estimate, our Bandwidth Calculator helps model expected served GB from historical origin logs before you flip the DNS.
Step‑by‑Step: Pulling the Right Metrics from Cloudflare
Finding True Cache Hit Ratio by Bandwidth
Log into the Cloudflare dashboard, select your zone, go to Caching → Cache Analytics. The default tile shows request‑based ratio. To get byte‑weighted, use the Traffic app and sum Cached vs Origin bandwidth over 30 days.
A first attempt trusted the 92% request hit ratio and projected $1,200 savings. Actual was $610 because video segments were dynamically generated and bypassed cache. Measure bytes, not requests.
Bandwidth Served vs. Requests
Cloudflare’s Traffic tab reports total bandwidth egress from its edge. That number is your denominator. If you’re on Free plan and can’t see long‑range data, approximate using server logs before activation, then refine after 30 days.
The Gotcha: Dynamic Content and Cache Bypass
API calls, personalized HTML, and admin paths often carry Cache‑Control: no‑store. Those bytes always hit origin. Your worksheet must subtract them. I recommend tagging such routes in nginx or using Cloudflare Cache Rules to quantify bypass explicitly.
Using Logpush for Audit‑Grade Accuracy
For enterprise cases, Cloudflare Logpush delivers hourly HTTP logs to R2 or S3. I once reconciled a $40k annual discrepancy by joining Logpush bytes with AWS VPC flow logs. The lesson: dashboards round; logs are exact.
Debunking the “Free Bandwidth” Myth and Cloudflare KV Costs
Search engines pair the query Is Cloudflare KV free? with bandwidth because KV is edge storage. The answer: KV has a free tier (e.g., 100,000 reads/day on Free) but beyond that you pay per million reads/writes and per GB stored. It is not “bandwidth” in the CDN sense, but if you use Workers to serve assets from KV, you create a new cost line that can partially offset CDN savings.
In a 2023 project, we moved 2 TB of static images to Workers KV for sub‑10ms edge delivery. The CDN egress savings were $180/month, but KV read charges hit $95 because of 30 million monthly requests. Net win, but not as clean as the myth suggests. See Cloudflare KV pricing for current rates.
This is why a serious calculation separates origin egress savings from new edge compute costs. They are different ledger lines, and CFOs notice if you blend them.
Also revisit the earlier question: Does Cloudflare charge for bandwidth? The CDN proxy, no. But KV, Workers, and some add‑ons do have usage fees. Conflating them is the root of most PAA confusion.
Bandwidth Alliance Partners: Stacking Extra Savings
The Bandwidth Alliance is a federation of storage and compute providers who waive egress to Cloudflare. If your origin is Backblaze B2, Wasabi, or similar, the bytes that miss cache travel to Cloudflare free of charge. That turns the “origin egress rate” in your worksheet to $0 for those assets.
Consider this comparison table:
| Scenario | Monthly Served GB | Origin $/GB | CHR | Net Origin Cost |
|---|---|---|---|---|
| No CDN (direct origin) | 10,000 | 0.09 | 0% | $900 |
| Cloudflare + AWS origin | 10,000 | 0.09 | 85% | $135 |
| Cloudflare + B2 (Alliance) | 10,000 | 0 (waived) | 85% | $0 |
The framework scales to multi‑origin setups. For estimating total transferred bytes before you have Cloudflare data, our Bandwidth Calculator helps model growth and seasonal spikes.
Be warned: Bandwidth Alliance only covers the segment from partner to Cloudflare, not last‑mile to user. Cloudflare still does not bill that last mile, but your provider relationship changes and you must verify the partner’s own fair‑use terms.
Edge Cases That Skew Your Savings Calculation
TLS Handshakes and Protocol Overhead
Cloudflare terminates TLS; your origin may have paid for TLS termination too. The saved compute is real but tiny. More importantly, HTTP/3 multiplexing can change byte counts versus your origin logs. Reconcile with a week of dual‑logging before trusting the worksheet.
SaaS Multi‑Tenant Origin Sharing
If you host many customers on one origin, a single Cloudflare zone may cache some tenants well and others poorly. Calculate per‑hostname CHR, not global, or you’ll overstate savings for the noisy tenant. I’ve seen a 70% global CHR hide a 10% CHR for an API‑heavy subdomain.
When Cloudflare Might Not Save You Money
For a low‑traffic brochure site with 5 GB/month and 30% CHR, moving to Cloudflare Free saves maybe $0.30. The administrative overhead dwarfs the gain. Conversely, a high‑miss API proxy might see worse latency and no egress win. Honest trade‑off: CDN caching helps static‑heavy workloads, not dynamic‑first architectures.
Compression and Brotli Effects
Cloudflare applies Brotli to compatible clients. Your origin might have sent gzip or uncompressed. The bytes served from edge are smaller, but origin egress (what you save) is based on what origin would have sent. Use origin‑side uncompressed size in the worksheet to avoid double‑counting compression gains.
Build Your Own Origin Egress Savings Spreadsheet
Create a Google Sheet with columns: A = Total Served GB, B = CHR (decimal), C = Origin $/GB, D = BA discount factor (0‑1). Row 2 formula: =A2*B2*C2*(1‑D2). That’s your monthly savings. Add a column for KV cost from Workers dashboard to net it out.
I keep a version with conditional formatting: if savings < $50, flag as “not worth engineering time.” That simple rule prevents wasted effort on small sites. The worksheet is a decision tool, not just a calculator.
For SMBs, I suggest a second tab that tracks three months of actual AWS/GCP egress line items versus projected. Variance > 15% means your CHR input is wrong—go back to Logpush.
A 30‑Day Measurement Plan to Prove the Number
Week 1: Export current origin egress from AWS Cost Explorer or GCP Billing. Establish baseline. Note any seasonal events.
Week 2: Enable Cloudflare, set cache rules, but don’t change app logic. Record Traffic bandwidth daily. Resist the urge to tune early; you want a clean before/after.
Week 3: Pull byte‑weighted CHR. Plug into worksheet. Compare projected vs actual origin bill. Expect a 5–10% gap due to dashboard rounding.
Week 4: Tune cache TTLs, activate Bandwidth Alliance if applicable, recalculate. I’ve found a second iteration typically lifts CHR by 5‑12 points, turning a $400 projection into $520 real.
Bottom line: calculating Cloudflare bandwidth savings is an exercise in origin accounting, not CDN invoicing. Master the worksheet and you can defend every dollar to your CFO.
Why This Approach Beats the Official Calculators
Cloudflare’s official Bandwidth Alliance page and Workers pricing calculator are useful, but they assume you already know your egress rate and CHR. They don’t teach you how to extract byte‑weighted CHR from a live zone. Our worksheet fills that gap with a reconciliation step most SMBs skip.
In practice, I’ve used both: the official tools for stakeholder slides, the worksheet for engineering forecasts. The two should match within 10%; if not, the worksheet is usually right because it uses your own billing data.
Remember the unique angle: Cloudflare offsets origin cloud fees rather than billing for bandwidth itself. Internalize that, and the PAA questions answer themselves.