Weighted A/B Testing — Custom Traffic Splits
How to use unequal traffic splits in your A/B tests and when a weighted split makes more sense than 50/50.
What is Weighted A/B Testing?
In a standard A/B test, traffic is split equally between variants — 50% to A and 50% to B. Weighted A/B testing lets you assign custom percentages instead, such as 80/20 or 90/10.
Weighted splits are available on the Premium plan and above. On the Pro plan, equal splits are supported.
When to Use an Unequal Split
Testing something risky — if Variant B is an experimental design you are not sure about, send only 10–20% of traffic to it. If it performs poorly, only a small fraction of your audience had a bad experience.
High-value campaigns — when your QR code is on premium packaging or a high-visibility placement, you want most traffic on the proven version while still gathering data on the new variant.
Gradual rollout — start at 90/10, then shift to 70/30, then 50/50 as confidence in the new variant grows. This is sometimes called a 'staged rollout'.
Validating a hypothesis quickly — if you are fairly confident Variant B will win, a heavier weight gives it more data faster for earlier statistical significance.
How to Set a Weighted Split
Weights do not need to be round numbers — you can use values like 67/33 if needed. Just ensure they add to 100.
Create or open an A/B test
Go to A/B Tests and create a new test, or open an existing one that has not yet started (or pause a running test to edit it).
Find the traffic weight fields
Each variant card shows a Weight input field, usually shown as a percentage.
Enter your desired split
Enter your percentage for each variant. All weights must add up to exactly 100.
Examples:
- Conservative test: A = 80, B = 20
- Lean toward new variant: A = 30, B = 70
- Three-way split: A = 60, B = 30, C = 10
Save and start
Click Save and then Start Test. SMLLR routes traffic proportionally to the weights you set from the very first scan.
Adjusting Weights on a Running Test
You can change variant weights while a test is running without ending it:
1. Open the active test
2. Click Edit Weights
3. Update the percentages
4. Save
The new split applies to all incoming scans from that moment. Historical scans are not retroactively redistributed — the analytics continue to accumulate separately per variant.
Changing weights mid-test can complicate interpretation of results. If you need to change weights significantly, consider ending the test, archiving the result, and starting a fresh test with the new weights.
How Weighted Splits Affect Your Needed Sample Size
With an unequal split, the minority variant takes longer to gather enough scans for reliable results. For example, with a 90/10 split:
- Variant A (90%) reaches 100 scans after ~111 total scans
- Variant B (10%) reaches 100 scans after ~1,000 total scans
If your QR code has low scan volume, consider a more balanced split to reach meaningful results faster. Weighted splits make the most sense for high-traffic QR codes where the minority variant still collects enough data quickly.
Aim for at least 100 scans per variant before drawing conclusions, regardless of split weight.
Was this article helpful?