A/B Testing

Test steps based on random traffic splits.

Easily test and compare different step designs inside your funnels. With Funnelish A/B Split Tests, you can create, manage, and optimize multiple step variants to see what converts best - all from one place. The results of your AB split test can be seen on the Funnel analytics page.

You can also set up variants for different countries (instead of random traffic) with a Geo split test.

Creating a variant step

  1. Open a funnel and select the step where you want to run the test.

  2. Click on New Variant.

  3. Choose how to create your variant:

    • Start from scratch – build a brand new version of the step.

    • Duplicate main variant – make a copy of your current step and tweak it.

    • Use a saved template – pick from your saved or Funnelish templates.

💡 Use the search bar to quickly find any variant you need. You can also filter by active or inactive variants for easier management.

New AB split test variant.

Managing Variants

  • Promote a variant → Make it your main step if it’s performing better.

  • Kill a variant → Remove underperforming ones without affecting the test flow.

  • Edit → Open and update the design at any time.

Kill/Promote Variant.

Setting Splits

You can switch between Random/Geo splits inside Step Settings:

Random/Geo Splits

Random Splits

  • Assign a percentage of traffic to each variant.

  • Click the field next to the variant name, choose a percentage, and repeat.

  • Don’t worry if totals don’t add up to 100% - Funnelish adjusts splits automatically based on relative values.

Tip: Keep it simple with 2–4 variants for best results.

Geo Splits

Want to test variants by country instead of random traffic?

  1. Go to the Step Settings.

  2. Switch from Random to Geo and save changes.

  3. Assign countries to each variant using the drop-down menu.

Viewing Results

Once your test is live, head to your Funnel Analytics page to track performance. Compare conversion rates, promote winners, and refine your funnel with data-driven decisions.

Last updated

Was this helpful?