Braze

Braze is a customer engagement platform represented in this wiki by its articles and whitepaper on multi-armed bandits, contextual bandits, and AI decisioning for lifecycle marketing.

Key facts

• Type: Company / customer engagement platform
• Source role: Publisher of "What is a multi-armed bandit? Smarter experimentation for real-time marketing" and the whitepaper "A community of bandits" [src-027, src-152]
• Key framing: Multi-armed bandits are a decisioning approach embedded in platforms such as Braze, while contextual bandits and a community-of-bandits architecture extend the pattern toward 1:1 customer decisions [src-027, src-152]
• Concepts introduced here: Intelligent Selection, Marketing Bandit Optimisation, Dynamic Traffic Allocation, Ab Testing Vs Bandits, AI Decisioning, Community Of Bandits [src-027, src-152]

What it does

The Braze article frames Multi Armed Bandits as a live campaign optimisation mechanism. Instead of waiting for a fixed A/B test to finish, a bandit continuously learns which message, offer, channel, timing, or creative variation is performing best and reallocates traffic while the campaign is still running ^[src-027].

Braze's product-specific layer is Intelligent Selection, described as part of Braze AI Decisioning Studio. In that framing, MAB algorithms handle the allocation problem, while contextual bandits and predictive scoring contribute broader personalisation and next-best-action decisions ^[src-027].

The source also uses Kayo Sports as an example of adaptive decisioning in customer engagement, with triggered campaigns improving conversion, message engagement, and churn outcomes through live message selection and timing ^[src-027].

The later Braze whitepaper deepens the implementation view. It argues that generative AI can help create content variants and support data work, but customer-level marketing still needs reinforcement-learning decisioning to choose the right action, timing, channel, and frequency for each person. Braze's named architecture is a Community Of Bandits, where multiple contextual bandits collaborate across decision dimensions rather than trying to optimise every possible campaign combination as one large model ^[src-152].

Related

• See also: Intelligent Selection, Marketing Bandit Optimisation, Multi Armed Bandits, Dynamic Traffic Allocation, Contextual Bandits, AI Decisioning, Ab Testing Vs Bandits, Community Of Bandits

Source references

• ^[src-027] Team Braze — "What is a multi-armed bandit? Smarter experimentation for real-time marketing" (2025-12-08)
• ^[src-152] George Khachatryan, Nathaniel Rounds, Victor Kostyuk / Braze – "A community of bandits" (source page: "From multivariate testing to AI decisioning", 2025-09-30)