A/B Test Full Traffic vs Partial Traffic Bandwidth Cost
Compare how full and partial test allocations, heavier and lighter variants, and two- versus three-version tests affect bandwidth estimates.
Bandwidth cost depends on how much traffic enters the experiment and how the weighted test page size compares with the control. These comparisons explain the main trade-offs without assuming a particular provider or pricing plan.
- 100% Free
- No Sign-Up Required
- Private & Secure
- Mobile Friendly
About A/B Test Full Traffic vs Partial Traffic Bandwidth Cost
Bandwidth cost depends on how much traffic enters the experiment and how the weighted test page size compares with the control. These comparisons explain the main trade-offs without assuming a particular provider or pricing plan.
3
Comparisons
5
Key Factors
Instant
Results
100%
Free to Use
Full-Traffic Test vs Partial-Traffic Test
The same page-weight difference produces different incremental transfer depending on the portion of traffic enrolled.
| Factor | Option A: 100% Test Traffic | Option B: 25% Test Traffic | What It Means |
|---|---|---|---|
| Test page views | All measured page views enter the test. | One quarter of measured page views enter the test. | The appropriate allocation depends on the experiment design and desired exposure. |
| Incremental bandwidth | Applies across the full measured audience. | Applies only to the enrolled quarter. | With identical page weights and duration, lower allocation reduces incremental transfer proportionally. |
| Total test-period bandwidth | More affected by the test page-weight average. | Mostly remains at the control page weight. | Non-test traffic is assumed to continue receiving the control. |
| Speed of reaching test sample size | Typically reaches a given traffic volume sooner. | Typically takes longer to collect the same number of test page views. | More enrolled traffic produces more experiment exposure per day. |
| Egress-cost exposure | Highest when variations are heavier than the control. | Lower for the same page-weight difference. | Only enrolled traffic is subject to the experimental page-weight mix. |
Partial allocation generally reduces incremental bandwidth cost, while full allocation makes the page-weight impact apply to more traffic.
Heavier Variant vs Lighter Variant
The direction of the cost change depends on the weighted average page size of the test mix.
| Factor | Option A: Heavier Variant | Option B: Lighter Variant | What It Means |
|---|---|---|---|
| Average test page weight | Above the control weight when split evenly. | Below the control weight when split evenly. | A lower average delivered page size requires fewer transferred bytes. |
| Additional data transfer | Positive relative to the control-only baseline. | Negative relative to the control-only baseline. | The estimate measures the difference from serving the control to every page view. |
| Estimated egress cost | Increases when egress has a positive per-GB price. | Decreases under the same pricing assumption. | Cost follows the transfer difference multiplied by the egress rate. |
| Experiment validity | Not determined by page weight alone. | Not determined by page weight alone. | Bandwidth estimates do not evaluate experiment quality or outcome validity. |
Page weight determines the direction of estimated bandwidth change: heavier test mixes add transfer, while lighter mixes can reduce it.
Two-Version A/B Test vs Three-Version A/B/C Test
Adding versions changes the equal-split weighting and can change the average test page weight.
| Factor | Option A: Two Versions | Option B: Three Versions | What It Means |
|---|---|---|---|
| Traffic split | 50% control and 50% one variation at full allocation. | About one third per version at full allocation. | The calculator assumes each included version receives an equal share. |
| Input needed | One variation's average page weight. | Average page weight across both variations. | Two-version testing is simpler to represent when variation sizes differ. |
| Average test weight | Average of control and one variation. | Average of control and two variation weights. | The result depends on the actual sizes of the versions. |
| Bandwidth predictability | Direct when one variation weight is known. | Requires a representative average variation weight. | A single average can obscure differences between multiple variants. |
| Incremental egress | Depends on the variation-control size gap. | Depends on the weighted average across both variations. | Version count alone does not determine bandwidth cost. |
More versions do not automatically cost more bandwidth; the key factor is the weighted average transfer size of all versions served in the test.
Key Differences at a Glance
Test traffic allocation controls how many page views are exposed to the experimental page-weight mix.
A heavier weighted test mix increases incremental transfer, while a lighter mix can reduce it.
Total versions affect the equal-split average page weight, not necessarily the direction of cost.
The total egress cost includes all measured traffic; additional cost isolates the difference from the all-control baseline.
A lower incremental bandwidth estimate does not measure experiment performance, conversion impact, or platform cost.
How to Decide
Assumptions
- Traffic assigned to a test is evenly split across all versions, including the control.
- All non-test page views receive the control page.
- Page transfer size remains consistent enough to use an average value.
- The same egress rate applies to baseline and test-period transfer.
- The comparison addresses bandwidth and egress only, not testing outcomes or other operating costs.
Related Comparisons
Frequently Asked Questions
Is partial traffic always cheaper for A/B test bandwidth?
For the same duration, traffic, and page-weight difference, partial allocation reduces the amount of traffic exposed to the test mix and therefore usually reduces incremental transfer.
Does adding a third version automatically increase egress cost?
No. The result depends on the average page weight across all versions. A third version can be lighter, similar, or heavier than the control.
Should I choose full traffic or partial traffic based only on bandwidth cost?
No. Bandwidth is one planning factor. The suitable experiment allocation also depends on goals and implementation considerations outside this calculation.
Why compare additional cost with total test cost?
Additional cost isolates the impact of page-weight changes, while total cost estimates all measured egress during the test period.
Ready to calculate your result?
Try the calculator and compare options with your own inputs.