Multi-Armed Bandits automatically find the best variant among a group of candidates by balancing between "exploring" options and "exploiting" the best option through dynamic traffic allocation. On Statsig, Bandits are used to select the best user experience to drive a target metric or action.

Statsig's [Autotune](./multi-armed-bandit) (the Multi-Armed Bandit solution) allocates traffic towards high-performing variants and can eventually identify a winning variant.

Statsig's [Autotune AI](./contextual-bandit) (the Contextual Bandit solution) is a personalization tool that serves users the best variant determined by a machine learning model trained on previous observations.

## How Autotune works

Autotune is Statsig's [Bayesian Multi-Armed Bandit](./multi-armed-bandit), and Autotune AI is Statsig's [Contextual Bandit](./contextual-bandit).

Both Autotune products test and measure different variations and their effect on a target metric.

* The multi-armed bandit continuously adjusts traffic toward the best-performing variations until it can select the best variation with confidence. The winning variation then receives 100% of traffic.
* The contextual bandit personalizes which variant a user sees based on provided user or interaction attributes, serving each user the variation predicted to be best for them.

Contextual Bandits are a subset of Multi-Armed Bandits. Both seek to balance the explore/exploit problem: choosing between exploiting the current best known solution and exploring to gather more information about other solutions.

The blog posts on [Multi-Armed Bandits](/autotune/multi-armed-bandit) and [Contextual Bandits](https://www.statsig.com/blog/statsig-autotune-contextual-bandits-personalization) cover use cases and considerations in depth. The table below summarizes the main considerations for when to use bandits, a ranking engine, or an experiment.

|                            | A/B/n Test | Multi-Armed Bandit (Autotune) | Contextual Bandit (Autotune AI)  | Ranking Engine                           |
| -------------------------- | ---------- | ----------------------------- | -------------------------------- | ---------------------------------------- |
| Typical # Variants         | 2-3        | 4-8                           | 4-8                              | Arbitrary #                              |
| Personalization Factor     | None       | None                          | Moderate                         | High                                     |
| Input Data Required        | None       | Very Little (100+ samples)    | Little - generally 1000+ samples | Tens of thousands to millions of samples |
| Model Efficacy             | None       | Basic                         | Moderate                         | High                                     |
| Identifies Best Variant    | Yes        | Yes                           | No                               | No                                       |
| Consistent User Assignment | Yes        | No                            | No                               | No                                       |