A/B Testing Bandwidth Cost Calculator Examples
Worked examples showing how page size, traffic allocation, and egress rates affect an A/B test bandwidth estimate.
These examples use representative traffic and page-weight scenarios to show how the calculator estimates added transfer. The examples are educational estimates and use decimal GB, with 1 GB equal to 1,000,000 KB.
Full-traffic A/B test with a heavier design
A marketing site tests a redesign with additional imagery and scripts.
Input Summary
Daily visitors
10,000
Page views per visitor
3
Duration
14 days
Test traffic
100%
Versions
2
Control / variant size
800 KB / 950 KB
Egress rate
$0.09 per GB
Calculation Breakdown
- 1Total page views10,000 * 14 * 3420,000
- 2Average test weight(800 + 950) / 2875 KB
- 3Extra transfer(420,000 * (875 - 800)) / 1,000,00031.5 GB
- 4Extra cost31.5 * 0.09$2.84
Result Summary
Total page views
420,000
A/B Testing Bandwidth Cost Calculator
Estimated total test transfer is 367.5 GB, compared with 336 GB for the control-only baseline.
Partial-traffic experiment on a high-traffic site
A product team tests a media-rich product-page layout for one month.
Input Summary
Daily visitors
250,000
Page views per visitor
2
Duration
30 days
Test traffic
20%
Versions
2
Control / variant size
600 KB / 1,000 KB
Egress rate
$0.08 per GB
Calculation Breakdown
- 1Total page views250,000 * 30 * 215,000,000
- 2Test page views15,000,000 * 0.203,000,000
- 3Extra transfer3,000,000 * ((600 + 1,000) / 2 - 600) / 1,000,000600 GB
- 4Extra cost600 * 0.08$48.00
Result Summary
Total page views
15,000,000
A/B Testing Bandwidth Cost Calculator
The test adds an estimated 600 GB of transfer, while total test-period transfer is about 9,600 GB.
Three-version test with a lighter average variation
A publisher tests two streamlined layouts against an older, heavier control.
Input Summary
Daily visitors
40,000
Page views per visitor
4
Duration
21 days
Test traffic
75%
Versions
3
Control / average variant size
1,200 KB / 900 KB
Egress rate
$0.10 per GB
Calculation Breakdown
- 1Total page views40,000 * 21 * 43,360,000
- 2Test page views3,360,000 * 0.752,520,000
- 3Average test weight(1,200 + (2 * 900)) / 31,000 KB
- 4Transfer change2,520,000 * (1,000 - 1,200) / 1,000,000-504 GB
- 5Cost change-504 * 0.10-$50.40
Result Summary
Total page views
3,360,000
A/B Testing Bandwidth Cost Calculator
Estimated test transfer is 3,528 GB versus a 4,032 GB control-only baseline.
How to Read Your Results
Additional data transfer is the difference between test-period transfer and the control-only baseline.
A positive additional cost means the test mix is heavier on average than the control.
A negative additional cost indicates that the test mix is lighter on average.
Total egress cost includes all measured page views during the period, not only the experimental variants.
Use representative delivered page weights rather than a single uncached page-load measurement.
Assumptions & Important Notes
- Versions receive equal shares of traffic included in the test.
- Page sizes are average KB transferred per measured page view.
- Egress is billed at one consistent rate per decimal GB.
- Traffic patterns do not change materially because of the experiment.
Related Examples
Frequently Asked Questions
What happens if only part of traffic is included in an A/B test?
Only the assigned test share uses the weighted average of the control and variants. The remaining traffic is calculated at the control page weight.
How do I estimate a page's transfer size?
Use CDN data, server logs, real-user monitoring, or browser network reports, then choose a representative average for the pages being measured.
Can I use this calculator for an A/B/C test?
Yes. Set total versions to 3, including one control and two variations, then enter the average page weight across the variations.
Does a lower page weight guarantee a lower bill?
Not necessarily. Included transfer, price tiers, cache behavior, regions, and other billed services can affect the final invoice.
Ready to calculate your own result?
Use the live calculator with your own inputs, timing, and preferences.