A/B Testing Bandwidth Cost Per-User Formula
Learn how to estimate the incremental bandwidth cost per variant user and total delivery cost for an A/B test.
This calculation estimates the extra data transfer and delivery cost created when a test variant is larger than the control. It separates the per-exposed-user impact from the full-test impact, helping teams assess the operational cost of heavier pages or assets during an experiment.
- 100% Free
- No Sign-Up Required
- Private & Secure
- Mobile Friendly
Extra Bandwidth Cost per Exposed User
Where:
Find how many more KB the variant transfers per view, multiply by the expected views for one variant user, convert KB to GB, and multiply by the bandwidth price per GB.
Variables Explained
| Variable | What It Means | Unit |
|---|---|---|
| variantSizeKb - Variant transfer size | Average data transferred when a user loads the variant experience. | KB per page view |
| controlSizeKb - Control transfer size | Average data transferred when a user loads the control experience. | KB per page view |
| pageViewsPerExposedUser - Page views per exposed user | Average number of tested-experience loads made by one user assigned to the variant. | number |
| bandwidthRate - Bandwidth rate | Data-delivery price in the selected currency for each GB transferred. | currency per GB |
| testUsers - Total test users | Expected number of unique users included in the experiment. | users |
| variantAllocation - Variant traffic allocation | Percentage of test users assigned to the variant. | percent |
Step-by-Step Calculation
Calculate users exposed to the variant
Only users assigned to the variant are included in the incremental transfer total.
variantUsers = testUsers * (variantAllocation / 100)
Find the extra transfer per page view
The calculation counts only a positive size increase. A same-size or smaller variant has zero additional transfer.
additionalKbPerView = max(variantSizeKb - controlSizeKb, 0)
Calculate extra transfer per exposed user
This applies the page-size difference across the expected number of tested page views.
additionalKbPerUser = additionalKbPerView * pageViewsPerExposedUser
Convert per-user transfer to GB
The calculator uses 1 GB = 1,048,576 KB.
additionalGbPerUser = additionalKbPerUser / 1048576
Calculate cost per exposed user
Multiply additional GB per variant user by the delivery rate.
bandwidthCostPerUser = additionalGbPerUser * bandwidthRate
Calculate full-test cost
Multiply the per-user incremental cost by the estimated number of users who receive the variant.
totalBandwidthCost = bandwidthCostPerUser * variantUsers
Example: 150 KB larger variant with 50% traffic allocation
Variant users
100000 × (50 / 100)
50,000 users
Extra transfer per view
max(400 − 250, 0)
150 KB
Extra data per exposed user
150 × 1
150 KB
Extra GB per exposed user
150 ÷ 1048576
0.000143051 GB
Extra cost per exposed user
0.000143051 × 0.08
0.000011444
Total additional cost
0.000011444 × 50000
0.5722
Final Result
The larger variant adds about 150 KB and 0.000011 in bandwidth cost per exposed user. Across 50,000 variant users, it adds about 7.15 GB of transfer and 0.57 in the selected currency.
Assumptions
- ✓Control and variant sizes are average transferred sizes for one page view or experience load.
- ✓The variant allocation is applied evenly to the expected test-user total.
- ✓Every variant user generates the entered average number of tested page views.
- ✓The bandwidth rate remains constant for all incremental GB in the experiment.
- ✓One GB is treated as 1,048,576 KB.
Limitations
- !Actual invoices may also include request, storage, edge-compute, cache, minimum-fee, tax, or regional-price components.
- !Caching, reused assets, compression, redirects, and protocol overhead can make measured transfer differ from the input size.
- !User behavior may vary by segment, device, geography, and experiment duration.
- !The calculator does not estimate performance effects, conversion effects, or infrastructure capacity costs.
Common Mistakes to Avoid
Entering uncompressed build size instead of the average compressed transfer size received by users.
Using all test users as variant users rather than applying the variant traffic allocation.
Entering total site page views instead of page views of the tested experience per variant user.
Using a bandwidth price in a different currency than the one used for reporting.
Treating a smaller variant as a negative cost when this calculator is designed to report additional cost only.
Related Formulas
Frequently Asked Questions
How is A/B testing bandwidth cost per user calculated?
Subtract the control transfer size from the variant size, cap negative differences at zero, multiply by views per exposed user, convert KB to GB, and apply the bandwidth rate.
Why does the formula use max(variant size minus control size, 0)?
The calculator measures added cost from a larger variant. If the variant is smaller, it reports zero additional cost rather than a negative value.
How many KB are in the GB conversion used here?
The calculation uses 1,048,576 KB per GB, which is 1024 × 1024 KB.
Does traffic allocation change the per-user cost?
No. Allocation changes the number of users receiving the variant and therefore the total cost, not the cost for one exposed user.
Should page views per user include reloads?
Include average loads of the tested experience that are expected to transfer the relevant variant assets. Use measured behavior where possible.
Ready to calculate your result?
Use the calculator to get instant results with your own inputs.