A/B Test Control vs Variant Bandwidth Cost
Compare control-only delivery with mixed A/B test delivery and see how page weight and traffic splits affect annual bandwidth cost.
A/B test delivery cost depends on the volume of eligible page views, the page weight of each experience, and the share of traffic served each one. These comparisons show when a mixed test costs more, less, or the same as a control-only baseline.
- 100% Free
- No Sign-Up Required
- Private & Secure
- Mobile Friendly
About A/B Test Control vs Variant Bandwidth Cost
A/B test delivery cost depends on the volume of eligible page views, the page weight of each experience, and the share of traffic served each one. These comparisons show when a mixed test costs more, less, or the same as a control-only baseline.
3
Comparisons
5
Key Factors
Instant
Results
100%
Free to Use
Equal-weight control and variant
Both experiences transfer the same average number of KB per page view.
| Factor | Option A: Control Only | Option B: 50/50 A/B Test | What It Means |
|---|---|---|---|
| Transferred page weight | 1,500 KB on every test view | 1,500 KB for both experiences | The weighted average page weight is identical. |
| Annual bandwidth | Based on all test views at 1,500 KB | Same as control-only baseline | Splitting traffic does not change total page-delivery transfer when weights match. |
| Incremental bandwidth cost | Baseline | Zero from page-weight difference | There is no delivery-cost change attributable to the variant weight. |
| Experiment learning | No comparison of experiences | Measures the selected test outcome across two experiences | This is a measurement consideration rather than a bandwidth result. |
With equal transferred weights, a traffic split does not increase page-delivery bandwidth under this model.
Heavier variant at a balanced split
The variant transfers more data and receives half of test traffic.
| Factor | Option A: Control Only | Option B: 50/50 Test With Heavier Variant | What It Means |
|---|---|---|---|
| Example page weight | 1,500 KB | 1,500 KB control and 1,700 KB variant | The test mix averages 1,600 KB per view, which is heavier than control. |
| Bandwidth per 100,000 test views | 150 GB | 160 GB | The heavier variant adds 10 GB per 100,000 test views in this example. |
| Annual delivery cost | Lower at the same CDN rate | Higher at the same CDN rate | Additional transfer is charged at the entered effective rate. |
| Ability to evaluate the variant | No experiment comparison | Supports comparison | The delivery increase may be acceptable depending on the experiment's purpose and duration. |
| Incremental cost signal | Baseline | Positive | A positive incremental result indicates more estimated delivery cost than control only. |
A heavier variant increases estimated delivery cost in proportion to its weight increase and share of test traffic.
Lighter variant with an 80% rollout share
The variant is optimized and receives most of the experiment audience.
| Factor | Option A: Control Only | Option B: 80% Lighter Variant Mix | What It Means |
|---|---|---|---|
| Example page weight | 2,400 KB | 2,400 KB control and 2,000 KB variant | The weighted average falls to 2,080 KB per test view. |
| Bandwidth per 100,000 test views | 240 GB | 208 GB | The mixed experience transfers 32 GB less per 100,000 views. |
| Annual delivery cost | Higher at the same traffic and rate | Lower at the same traffic and rate | Lower transfer produces a lower modeled data-transfer cost. |
| Incremental cost signal | Baseline | Negative | A negative incremental result indicates an estimated reduction versus control-only delivery. |
| Sensitivity to split | No split variable | Savings rise as more traffic receives the lighter variant | The variant share directly affects the weighted average page weight. |
A lighter variant can lower estimated annual delivery cost, especially when it serves a large share of test traffic.
Key Differences at a Glance
Control-only delivery uses one page weight for every test-eligible view, while an A/B test uses a weighted average of two weights.
The variant split affects only the test-eligible traffic, not total site traffic.
Equal page weights produce the same modeled bandwidth regardless of split.
A heavier variant creates a positive incremental cost; a lighter variant can create a negative incremental cost.
Traffic volume and pages per visitor scale both baseline and test costs proportionally.
How to Decide
Assumptions
- Comparisons use the same monthly visitor volume, pages per visitor, test allocation, and CDN rate for both options.
- One GB is treated as 1,000,000 KB.
- Each page view receives either control or variant, not both.
- Only page-delivery data transfer is compared; platform and operational costs are excluded.
Related Comparisons
Frequently Asked Questions
Is a control-only baseline the same as the site's total CDN bill?
No. It represents only the estimated traffic included in the experiment, priced as though it all received the control page.
When does a 50/50 A/B test cost the same as control only?
When control and variant have the same average transferred page weight.
How does increasing variant allocation affect cost?
It moves the weighted average page weight toward the variant. Cost rises for a heavier variant and falls for a lighter one.
Can this comparison choose the best experiment design?
It compares modeled delivery costs only. Other factors, such as experiment goals and user experience, are outside the calculation.
Ready to calculate your result?
Try the calculator and compare options with your own inputs.