A/B Testing

Last updated: 2026-03-27
Axerto's A/B testing engine lets you split-test multiple variants of a landing page. Traffic is allocated by weighted random sampling, with 30-day sticky cookies so each visitor always sees the same variant. Statistical significance is calculated automatically using z-tests at 90%, 95%, and 99% confidence. Enable Auto-Optimize to let the Bayesian algorithm automatically shift traffic toward the best-performing variant — similar to Unbounce Smart Traffic, included at no extra cost.

How to use it

  1. Create a variant

    Open the A/B Testing tab on your landing page. Click 'New Variant'. A copy of the current page is created as a new variant.

  2. Edit the variant

    Click 'Edit Content' on the variant to open it in the Visual Editor. Change the headline, CTA, hero image, or layout — anything you want to test.

  3. Set traffic weights

    Adjust the percentage for each variant. Weights are normalized to 100%. Example: 70/30 sends 70% of traffic to the original and 30% to the variant.

  4. Activate the variant

    Toggle the variant to Active. It will now receive traffic according to its weight. Deactivate at any time without losing data.

  5. Enable Auto-Optimize (optional)

    Toggle 'Auto-Optimize' in the A/B Testing header. Axerto uses a Bayesian multi-armed bandit algorithm to automatically shift traffic toward the highest-converting variant every 4 hours. A minimum traffic threshold (default 100 views per variant) ensures the decision is data-driven.

  6. Monitor results

    The A/B Testing tab shows views, leads, and conversion rate per variant. Significance levels are calculated automatically. When Auto-Optimize is on, the Rebalance History section shows every traffic shift with before/after weights.

  7. Promote the winner

    When a variant reaches statistical significance, click 'Promote'. It becomes the new main page; all other variants are deactivated.

Frequently asked questions

How does Axerto assign visitors to variants?
On the first visit, a weighted random selection picks a variant. The result is stored in a 30-day cookie so the same visitor always sees the same version.
What is Auto-Optimize?
A Bayesian multi-armed bandit algorithm that automatically adjusts traffic weights every 4 hours based on each variant's conversion rate. It uses Laplace-smoothed posterior means and a 5% exploration floor so no variant is ever completely cut off. When weights change by less than 2%, no rebalance is triggered to avoid noise.
What significance threshold should I use?
95% (p<0.05) is the industry standard for most marketing tests. Use 99% for high-stakes decisions. 90% is acceptable for early exploratory tests with limited traffic.
Can I test more than 2 variants simultaneously?
Yes. There is no hard limit on variant count. Be aware that more variants require proportionally more traffic to reach statistical significance.
Is historical A/B data preserved after promoting a winner?
Yes. View count and conversion data are permanently stored on each Landing record.
Related
Visual Editor Conversion Analytics AI Conversion Suggestions