A/B Testing Website Load Time Per-User Formula
Learn how to calculate per-user load-time savings, percentage change, traffic-weighted test load time, and cumulative impact in an A/B test.
This calculator compares a control page with a variant to estimate how much waiting time each variant user saves or gains across their measured page loads. It also scales that difference by variant exposure, which helps describe the operational size of a performance change.
- 100% Free
- No Sign-Up Required
- Private & Secure
- Mobile Friendly
Time Saved Per Variant User
Where:
Subtract the variant load time from the control load time, then multiply by the average number of measured page loads per user. A positive result means the variant is faster; a negative result means it is slower.
Variables Explained
| Variable | What It Means | Unit |
|---|---|---|
| controlLoadTime - Control load time | Average measured load time for the control version. | seconds |
| variantLoadTime - Variant load time | Average measured load time for the test variant. | seconds |
| pageLoadsPerUser - Average page loads per user | Average number of measured loads experienced by each user. | number |
| totalUsers - Total test users | Total number of users included in the test period. | number |
| variantTrafficShare - Variant traffic share | Percentage of total test users assigned to the variant. | percent |
| variantUsers - Variant users | Estimated number of users exposed to the variant. | number |
Step-by-Step Calculation
Estimate users assigned to the variant
Convert the traffic share to a decimal and multiply it by total test users.
variantUsers = totalUsers * (variantTrafficShare / 100)
Find the load-time difference per page load
A positive difference indicates that the variant is faster for one measured load.
loadTimeDifference = controlLoadTime - variantLoadTime
Calculate time saved per user
Multiply the per-load difference by the number of relevant page loads per user.
timeSavedPerUser = loadTimeDifference * pageLoadsPerUser
Calculate percentage load-time change
This expresses the variant's change relative to the control load time.
loadTimeChangePercent = ((controlLoadTime - variantLoadTime) / controlLoadTime) * 100
Calculate the traffic-weighted average test load time
Combine control and variant load times according to their respective traffic shares.
averageTestLoadTime = (controlLoadTime * (1 - variantTrafficShare / 100)) + (variantLoadTime * (variantTrafficShare / 100))
Calculate cumulative time saved for variant users
Scale the per-user result by the estimated number of users who saw the variant.
totalTimeSaved = timeSavedPerUser * variantUsers
Example: Faster variant with two measured loads per user
Variant users
10,000 * (50 / 100)
5,000 users
Difference per page load
3.0 - 2.4
0.60 seconds
Time saved per user
0.60 * 2
1.20 seconds
Load-time change
((3.0 - 2.4) / 3.0) * 100
20.0% faster
Average test load time
(3.0 * 0.5) + (2.4 * 0.5)
2.70 seconds
Total time saved
1.20 * 5,000
6,000 seconds
Final Result
The variant saves an estimated 1.20 seconds per user and 6,000 seconds across 5,000 variant users.
Assumptions
- ✓Control and variant use the same page, performance metric, device mix, and measurement period.
- ✓The entered values are representative averages for the users included in the test.
- ✓Each variant user experiences the entered average number of measured page loads.
- ✓The traffic allocation reflects the actual variant exposure during the test.
- ✓Positive time saved represents a faster variant, not a proven change in business outcomes.
Limitations
- !Average load time can hide important differences by device, network, geography, or user segment.
- !The calculation does not test statistical significance or quantify uncertainty.
- !It does not measure conversion, engagement, revenue, retention, or other behavioral outcomes.
- !A traffic-weighted average describes the test mix and is not necessarily the expected result after a full rollout.
- !Repeated loads, caching behavior, and navigation paths may differ from the entered average.
Common Mistakes to Avoid
Comparing different performance metrics, such as control page-load time against variant LCP.
Entering a total user count that includes traffic outside the measured test period.
Using the total number of site pageviews instead of measured loads for the tested page.
Treating a positive time-saved estimate as proof that the variant improves conversion.
Forgetting that a 50% traffic split means only half of total test users contribute to total variant savings.
Related Formulas
Frequently Asked Questions
What is the formula for time saved per user in a load-time A/B test?
Subtract variant load time from control load time and multiply the result by average measured page loads per user.
How is percentage load-time change calculated?
Divide the control-minus-variant difference by control load time, then multiply by 100.
What does a negative time saved per user mean?
It means the variant is slower than the control across the entered number of page loads per user.
How is total time saved calculated for variant users?
Multiply time saved per user by total users multiplied by the variant traffic share as a decimal.
Why does the calculator need traffic share?
Traffic share estimates how many people saw the variant, which is necessary to estimate cumulative waiting-time change.
Ready to calculate your result?
Use the calculator to get instant results with your own inputs.