Definition
A machine learning approach that constantly explores new options while exploiting what currently works—basically A/B testing on steroids with mathematical justification.
Example Usage
We implemented a bandit algorithm to optimize ad creative, and it's learning faster than our creative team can argue about fonts.
Origin
From the 'multi-armed bandit' problem in statistics
Fun Fact
Named after the idea of a gambler facing multiple slot machines, trying to maximize winnings—very on-brand for marketing optimization
Source: Machine learning and programmatic advertising
Related Terms
Translate This Term
See “Bandit algorithm” in Corporate Speak, Gen-Z Slang, Pirate Speak, and more.
Try the Translator