🧪 What is the Experiment Component?
The Experiment (A/B Testing) component in FlowBuilder allows you to randomly distribute your customer traffic across different configured routes (variants). Its main objective is to help you compare results natively, without needing to build complex technical solutions (such as having to use HTTP requests and conditionals).
Main benefits:
Direct distribution: Divide traffic by percentage to validate messages, rules, or experiences.
Zero calculation errors: Percentages auto-adjust in real time so they always sum to exactly 100%.
Agile iteration: Create, adjust, and delete variants visually from the canvas.
⚙️ Step-by-Step Configuration
You'll find this tool in the left side menu of FlowBuilder components, under the name Experiment.
To configure it:
Add the component: Drag it onto your canvas. When first inserted, it will be created by default with two outputs: Variant A (50%) and Variant B (50%).
Open the inspector: Double-click on the node. In the right side panel you'll be able to view and edit the experiment name to identify it easily.
Edit the variants: In the "Variants" section, click on each card to expand it. You'll be able to change the variant name and adjust the exact percentage you want to assign to it.
Add new routes: Click on ADD VARIANT if you need to test more than two paths. The system will create a new output on the canvas (Variant C, D, etc.) and redistribute the percentages equitably.
Delete variants: If you change your mind, you can delete cards from the inspector. The output port will disappear from the canvas and the remaining percentages will adjust automatically.
⚖️ Percentage Rules and Limits
To ensure your traffic division works without errors, the component has intelligent logic:
Auto-adjust (Always 100%): If you have two variants (50/50) and change "A" to 70%, the system will automatically adjust "B" to 30%. If you have three (40/30/30) and raise the first to 60%, the other two will adjust proportionally to share the remaining 40%.
Minimum variants (2): If you try to delete a variant and only two remain, the delete button will be completely disabled to protect the meaning of the A/B test.
Maximum variants (10): You can have up to ten different test paths in a single experiment block.
🎯 Use Cases and Best Practices
Not sure where to start testing? Here are some ideas and recommendations to get the most out of your experiments:
What to test?: Compare persuasive texts (copy), evaluate different instructions (prompts) in your Artificial Intelligence nodes (Smartons), or compare the effectiveness between a short flow and a long one.
Define your success metric: Before launching the experiment, be clear about what you're going to measure (conversion rate, response percentage, resolution time, etc.).
Don't interrupt it: Avoid changing percentages or variant names while the test is running live, as this will mix your data and ruin your ability to compare them.
Keep it simple: If you need quick results and a clear comparison, keep only 2 variants with a classic split (50/50 or 70/30).
❓ Frequently Asked Questions
Does the same customer always fall into the same variant? No. Currently, assignment is purely random with each interaction. If the contact completes the flow and passes through the component again later, they could fall into a different route.
Can I have more than 2 variants? Yes, you can configure from 2 to a maximum of 10 simultaneous variants in a single node.
What happens if I try to delete and only 2 variants would remain? The deletion is blocked in the interface, since the system requires a minimum of 2 options for a traffic division to exist.
Stop guessing and start measuring! 🚀
Using A/B Testing will allow you to make decisions based on real data, iterating your automation flows to always offer the experience and messages that convert best. ✅