Traffic allocation is the proportion of your visitors assigned to each variation in an A/B test. In a standard two-way test, 50% of eligible visitors see the control and 50% see the variant. Traffic allocation is set before the test launches and determines how quickly the test accumulates sample size — higher allocation means faster data collection, while lower allocation reduces the risk exposure of showing an unproven variant to your audience.
Estimated days to significance = Required sample size per variant / (Total daily visitors × Allocation % per variant)
Example: A test needs 15,000 visitors per variant. Your page gets 3,000 visitors per day.
- At 50/50 split (1,500 per variant/day): 15,000 / 1,500 = 10 days
- At 20/80 split (600 per variant/day): 15,000 / 600 = 25 days
Why Traffic Allocation Matters for Ecommerce
For D2C brands with limited daily traffic — common for stores in the ₹10–50 lakh monthly revenue range — traffic allocation directly determines how many experiments you can run concurrently and how quickly you get results. Setting allocation too low on a low-traffic store means tests take months to conclude, freezing your optimization roadmap. Setting it too high on an untested, risky variant means a large portion of your revenue-generating traffic is exposed to a potentially harmful experience. Getting this balance right is a core experimentation skill.
Real-World Example
Boat Lifestyle's CRO team wanted to test a major homepage redesign — a high-risk, high-reward variant. Rather than jumping to 50/50, they started with a 10/90 split (10% variant, 90% control) for the first five days to monitor error rates, bounce rate anomalies, and revenue-per-visitor at low exposure. Seeing no negative signals, they scaled to 50/50 for the full test duration. This staged ramp approach protected the majority of their traffic while still generating learnings quickly.
How to Set Traffic Allocation
- Default to 50/50 for standard tests — it minimizes test duration and produces balanced sample sizes.
- Use a phased ramp (10% → 50%) for major redesigns or changes to high-revenue pages where a bad variant could cause significant damage.
- Avoid allocating less than 10% per variant unless you are in a pure monitoring phase — below 10%, tests take impractically long to reach significance on most stores.
- Keep allocation constant throughout the test — changing it mid-test invalidates results by introducing a time-confound.
- When running multiple concurrent tests, ensure the combined allocation doesn't exceed 100% of eligible traffic per page or funnel step.
Traffic Allocation in A/B Testing
Traffic allocation feeds directly into run time calculations. Most testing platforms allow you to set the allocation split when configuring an experiment, and some offer automated allocation (multi-armed bandit) which dynamically shifts traffic to the better-performing variant. For most ecommerce tests, fixed equal-split allocation remains the most statistically sound approach.
Run smarter A/B tests with CustomFit.ai — 14-day free trial, no credit card required.