Braze Interview Questions & Guide

A Campaign is a single-step (or single-channel) send — good for a one-off blast or a simple triggered message. A Canvas is a multi-step, multi-channel journey with branching, delays, and audience paths. Strong candidates say they default to Canvas for anything with logic or sequencing, and reserve Campaigns for genuinely simple sends — and they mention Canvas Flow as the current builder.

A segment is a reusable, named audience defined by filters on attributes, events, and engagement; it updates continuously as users qualify or drop out. Inside a Canvas you can additionally apply entry filters and audience paths. The signal: knowing that segments are reusable and live, and that over-segmenting creates maintenance debt.

A custom attribute describes a state of the user (plan tier, lifetime value, city); a custom event records that something happened at a point in time (purchased, completed onboarding). Strong answers note that events can carry properties, that you trigger Canvases off events, and that modelling something as the wrong type (event vs attribute) causes pain later.

Subscription groups let users opt in/out of specific message types (e.g. promotions vs transactional) per channel, rather than a single global on/off. The mature answer ties this to compliance and to respecting preferences as a deliverability and trust lever — not just a legal checkbox.

Look for: an event-triggered entry, channel steps (push/email/in-app) with action paths and delays, fallbacks when a user has no push permission, and an exit/conversion event. The strong candidate sequences channels by cost and intrusiveness (in-app/email before SMS) and builds in a goal so the journey stops when the user activates.

Strong answers describe channel prioritisation logic, decision splits on channel reachability, and treating cross-channel coordination as a design problem — plus global frequency capping as a backstop. They distinguish ‘which channel’ from ‘how often’.

Frequency caps limit how many messages a user receives over a rolling window, globally or per channel/tag. The senior signal: knowing that caps are a safety net, not a strategy, that transactional/critical messages can be exempted, and that caps interact with Canvas priority.

Intelligent Timing sends to each user at the time they’re individually most likely to engage, based on their historical behaviour, instead of one fixed send time. Strong candidates note it’s most useful for non-time-sensitive sends and that they’d A/B test it against a fixed time rather than assuming a lift.

Look for comfort with Liquid attribute and event/property references, conditional logic ({% if %}), and crucially default values so a missing attribute doesn’t render an empty or broken message. The best answers stress testing across multiple user profiles.

Connected Content pulls data from an external API at send time and renders it into the message. Use it when data is too volatile or large to sync as attributes (live prices, inventory, weather). The mature answer flags latency/availability risk and the need for fallbacks if the endpoint fails.

Catalogs store structured reference data (products, content, locations) inside Braze that you reference in Liquid, avoiding an external call per send. Strong candidates contrast Catalogs (semi-static reference data) with Connected Content (live external data) and pick based on volatility and volume.

Via the SDKs (client-side events/attributes), the REST API / server-to-server (/users/track), and partner integrations (CDP, Segment, etc.). The signal: knowing which path suits which data, that server-side is more reliable for revenue/critical events, and that client and server data must be reconciled on a consistent identifier.

Currents is the streaming data export — it pushes engagement and message events to a data warehouse or analytics tool in near real time. Strong candidates use it to close the loop on attribution and build downstream reporting Braze’s native dashboards can’t.

Look for understanding of the external_id as the durable identity, anonymous users created by the SDK before login, and user aliasing / identify calls to merge an anonymous profile into a known one. Getting this wrong causes duplicate profiles and broken targeting — a great candidate raises that risk unprompted.

Strong answers use a variant/experiment step or message variants, define a single primary conversion event up front, ensure adequate sample size, and read results against conversion — not opens. They avoid testing five things at once and know to let the test reach significance.

A holdout withholds messages from a random slice of the audience so you can measure the incremental impact of a campaign or the whole program, not just in-message engagement. The senior signal: insisting on a global holdout to prove lifecycle messaging actually drives revenue rather than taking credit for organic behaviour.

Look for: tightening segmentation and triggers so messages are more relevant, frequency capping, Intelligent Timing, subscription-group hygiene, sunsetting unengaged users from high-frequency journeys, and validating with a holdout. Methodical diagnosis over ‘send more’.