
A/B Testing Bandwidth Cost Calculator Examples
Worked examples showing how traffic, page weight, test allocation, and CDN pricing affect annual A/B test bandwidth cost.
These scenarios illustrate annual A/B testing bandwidth estimates using different traffic volumes, page-weight differences, test allocations, and traffic splits. Dollar amounts use the stated example CDN rate only.
Small test with equal page weights
A product team tests copy changes on 20% of a site with modest traffic.
Input Summary
Monthly visitors
25,000
Pages per visitor
2
Test allocation
20%
Control and variant page weight
900 KB each
Variant split
50%
CDN cost
$0.10 per GB
Calculation Breakdown
- 1Monthly test views25,000 × 2 × 20%10,000 views
- 2Monthly bandwidth10,000 × 900 ÷ 1,000,0009 GB
- 3Annual cost9 × 12 × $0.10$10.80
Result Summary
Annual cost
$10.80
A/B Testing Bandwidth Cost Calculator
Estimated annual transfer is 108 GB and estimated annual cost is $10.80.
Higher-traffic 50/50 test with a heavier variant
A marketing site exposes half of its traffic to a 50/50 A/B test.
Input Summary
Monthly visitors
100,000
Pages per visitor
3
Test allocation
50%
Control page weight
1,500 KB
Variant page weight
1,700 KB
Variant split
50%
CDN cost
$0.08 per GB
Calculation Breakdown
- 1Monthly test views100,000 × 3 × 50%150,000 views
- 2Monthly bandwidth((75,000 × 1,500) + (75,000 × 1,700)) ÷ 1,000,000240 GB
- 3Annual cost240 × 12 × $0.08$230.40
- 4Incremental cost$230.40 - $216.00$14.40
Result Summary
Incremental cost
$14.40
A/B Testing Bandwidth Cost Calculator
Estimated annual test transfer is 2,880 GB, costing $230.40, with a $14.40 increase versus control only.
Variant-heavy rollout with a lighter page
An ecommerce team sends 80% of experiment traffic to a lighter product-page design.
Input Summary
Monthly visitors
500,000
Pages per visitor
4
Test allocation
40%
Control page weight
2,400 KB
Variant page weight
2,000 KB
Variant split
80%
CDN cost
$0.06 per GB
Calculation Breakdown
- 1Monthly test views500,000 × 4 × 40%800,000 views
- 2Monthly bandwidth((160,000 × 2,400) + (640,000 × 2,000)) ÷ 1,000,0001,664 GB
- 3Annual test cost1,664 × 12 × $0.06$1,198.08
- 4Control-only baseline800,000 × 2,400 ÷ 1,000,000 × 12 × $0.06$1,382.40
Result Summary
Control-only baseline
$1,382.40
A/B Testing Bandwidth Cost Calculator
Estimated annual transfer is 19,968 GB and annual cost is $1,198.08, which is $184.32 below the control-only baseline.
How to Read Your Results
Annual test bandwidth is the estimated yearly transfer for page views included in the experiment, not all site traffic.
Annual test cost applies the entered effective CDN price to that estimated transfer volume.
A positive incremental cost means the experiment mix is heavier than serving control only.
A negative incremental cost means the variant mix is lighter than the control baseline.
Use the figures for planning comparisons; validate them against provider usage reports and measured transfer sizes.
Assumptions & Important Notes
- Each example uses decimal GB, where 1 GB equals 1,000,000 KB.
- Traffic and page-view behavior are assumed to remain consistent every month.
- The stated CDN rate is treated as a single effective price.
- Page weights represent transferred page weight per relevant view.
Related Examples
Frequently Asked Questions
Should I use unique visitors or sessions in these examples?
Use the same visitor measure that supports your pages-per-visitor estimate. The key is producing a realistic page-view estimate.
What if my test runs for less than a year?
Calculate the annual estimate as a planning figure, then scale it approximately by the number of months the test will run.
Can I enter a 90/10 variant split?
Yes. The variant percentage is applied to test traffic, and the remaining test traffic is assigned to control.
Why might my invoice differ from an example?
Actual billing can reflect cache behavior, pricing tiers, geographic distribution, included transfer, and services not modeled here.
Ready to calculate your own result?
Use the live calculator with your own inputs, timing, and preferences.