Multi-Armed Bandit Distribution (May 2024)

Automatically adjusting the variant distribution to the winning variant

Today, we are releasing even-split and multi-armed bandit distribution types for Ninetailed Experiments, making it easier to automate the distribution of traffic across your experiences.

The multi-armed bandit approach takes its name from a hypothetical gambling scenario, where a player seeks to maximize returns by choosing the most rewarding slot machine (or "arm") among several.

In the realm of AI and experimentation, MAB algorithms apply this principle to manage and allocate resources between competing strategies based on real-time performance data.

Unlike traditional methods, which divide resources equally regardless of outcome, MAB continuously adjusts its strategy to favor the most effective options. This smart allocation leads to more efficient and effective experimentation, optimizing every interaction with your audience.

To find out more about multi-armed bandit distribution, have a look at our Experiments section.