Find out which customer service sells more with the A/B Test
Run two or more versions of your customer service with real customers at the same time, and compare on the Experiments page which one resolves more, hands off less to your team and brings in more sales.
Opinions are not results
A shorter message or a complete one? The offer up front or after the question? A consultative AI agent or a direct one? Without measuring, the choice is whoever built the flow’s guess.
The A/B Test splits contacts between the versions and measures what happened to each group, from the end of the conversation to the sale.
How to set up a test
Add the A/B Test block
In the flow editor, connect each variant to a different path and set each one’s percentage: 50/50, 70/30 or up to 10 variants.
Mark the goal (optional)
At the point that counts as success, such as a qualified lead or a booking, add the Track event block. That event becomes the test goal.
Compare in Experiments
In Flows → Experiments, pick the period and the goal and see the variants side by side.
What you compare
Each contact in one version only
A contact is drawn once and always stays in the same variant, even when they come back days later. Nobody sees both versions, and results don’t mix.
Conversation outcome
How many finished the flow, how many the AI resolved on its own, how many went to your team and how many stopped midway.
Goal and sales
How many reached the goal, how many bought in the Store (paid or delivered orders) and how many became a won CRM opportunity, with the amounts.
AI cost per variant
When the test compares AI agents, the page shows how much each variant consumed, so you can weigh results against cost.
Small-sample warning
With fewer than 30 participants per variant, the page warns that the difference may still be chance.
What it looks like in Wazapi
Who it is for
- Anyone who wants to choose messages, offers and flow paths based on numbers, not guesses.
- Teams using the AI Agent who want to compare instructions or different agents.
- Stores that want to know which service takes more customers to a paid order.
Frequently asked questions
Can the same customer see both versions?
No. The first time they reach the test they are drawn into one variant and stay there, even if they come back weeks later.
Can I test more than two versions?
Yes, up to 10 variants per block. The percentages must add up to 100%.
Can I change the percentages mid-test?
You can, but avoid it. Contacts who already joined keep their variant; only new contacts follow the new split.
Does the editor simulator count in the results?
No. In the simulator each run draws a variant on the spot without saving anything, so you can experiment freely.
Do I need the AI Agent to use the A/B Test?
No. The A/B Test works with any flow path: different messages, buttons, offers or schedules.
Want to see it working for your business?
Free 14-day trial, no credit card required.