Randomizer is a predefined branching block, used to perform **A/B Testing** or **Split Coupon**.

This feature allows you to set up a Randomizer Step inside your Flow. So you’ll be able to test anything inside your chatbots, from different button copies to entirely new message Flows…

As a result, you’ll be able to optimize and improve the performance of your chatbots over time.

To create the Randomizer block in the flow, just double tap/right mouse click anywhere in the flow's background and choose the needed block.

When the block is created, you will be able to choose the percentage of subscribers who will get each of the tested variants.

When the distribution is set up, you have to choose the steps the subscribers will follow. It can be either a message, an action, a flow, a condition, another Randomizer or a Smart Delay.

Please note: once subscribers reach Randomizer block with the "Random path every time" checkbox unticked, they are sorted into either A or B group. From there, it won't be possible to change the group that certain subscriber belongs to. So each time they receive the same flow again, they will be sent the exact same variant as before.

For large number of subscribers, with that setting, the Randomizer Block sends users down the specific path in parts. For example, you've send a message to 1000 people: Randomizer sends a certain amount of users down the first path. Then, it analyzes the number of users and sends other batches of users to other paths to make percentage even. 200 to A, 200 to B, 200 to C, 200 to A, 200 to B... and that's it. A will receive 400, B will receive 400 too and C only 200. That's why you can see this difference from time to time.

Results may vary and the difference may be bigger due to the sheer number of subscribers. That's another reason why we'd recommend Randomizer for ongoing Flows such as quizzes, sales, tournaments, so the Randomizer would always have a new group to send through and even the percentage.

If you are sending a one-time broadcast, we'd recommend using the "Random path every time" option: then, according to the law of large numbers, there will be no such difference in numbers because the Randomizer will be sending every user down the random path without checking statistics, and all users will be naturally distributed between all paths.