.svg)
Reducing churn with Braze comes down to four Canvas decisions: what starts the journey, how long it waits, what removes a user from it, and how you measure the result. Entry, delay, exit criteria and Action Paths each have documented behaviour that changes when a retention message sends. The order you configure them in matters.
Key Takeaways
- Design the exit before the first message, so every later choice follows from what a retained user actually does.
- A user in a Delay step who meets the exit criteria is marked to exit at once but leaves only when the delay ends.
- The Total Exits by Exit Criteria figure only updates after a user has fully exited the Canvas, so exit counts lag.
- Exit criteria end a journey, while Action Paths sort users onto branches and keep them in the Canvas.
- Add a control group, and set the conversion window to at least match the longest path in the Canvas.
- Stopping a Canvas does not exit users waiting in a Delay step, and those whose delay has passed exit instead of receiving the next step.
How does retention automation work in Braze?
Retention automation in Braze is a Canvas that reacts to a signal, waits, checks what the user did, and then decides what to send. Each of those moves is a Canvas component with documented behaviour. The craft is in choosing the behaviour you want at each one.
As a Braze implementation partner, we design the exit before the first message. That order forces an early answer to what a retained user actually does, and every later choice follows from it.
The Braze documentation defines no churn threshold and no lapse window. That definition belongs to your business, and it should be written down before anyone opens the Canvas editor.
Which entry trigger should start a retention Canvas?
You can choose one of three ways for users to enter a Canvas: scheduled, action-based or API-triggered. Scheduled entry runs on a time schedule, much like a campaign. Action-based entry starts when a user takes a particular action, such as opening your app, making a purchase or triggering a custom event.
Churn is usually an absence, and an absence can't fire an action-based trigger. For that reason a lapse sweep suits scheduled entry against a segment, while a journey that follows a specific event, such as a paused subscription, suits action-based entry.
Whichever you choose, the Target Audience step sets the gate. Braze evaluates the target audience for eligibility first, before users enter the Canvas journey.
The documentation also warns against using the same trigger as an audience filter, such as a changed attribute or a custom event the user performs. A race condition may occur in which the user is not in the audience at the time they perform the trigger event, so they never enter the Canvas. Entry controls then decide whether users may re-enter, which matters for anyone who can lapse more than once.
How long should a retention Canvas wait before it messages?
A Delay step holds users for a set time before the next step. Delay steps can run up to two years (730 days), so the platform leaves plenty of room. The business logic sets the real limit.
Profile changes made during a delay are picked up when the user moves on. Any changes to user attributes or contact information during the delay are reflected when the user advances to subsequent steps. A user who adds an email address while waiting can therefore receive the email step.
Braze calculates a Canvas's maximum duration by adding the delays and trigger windows along the longest path. Keep that figure in view, because it also tells you how long your conversion window should be.
Send settings deserve a place in the design too. Rate limiting and frequency capping ease the marketing pressure placed on users, which helps you avoid over-messaging them. You can also set quiet hours in the user's local time and choose whether a suppressed message aborts or sends at the next available time.
How do exit criteria stop retention messages at the right time?
Exit criteria remove users who have done the thing you wanted. When a user performs the exit event, Braze immediately marks them to exit the Canvas. They don't advance to any later step.
A user waiting in a Delay step behaves slightly differently. They remain in that step until the delay period ends, then they leave rather than moving on. The documentation's own example is a user in a 30 day delay who performs the exit event on day one: they're marked to exit right away but don't fully leave until the Delay step ends, 29 days later.
That behaviour has two consequences worth planning for.
First, exit counts lag. The "Total Exits by Exit Criteria" figure in Canvas analytics only updates after a user has fully exited the Canvas, so a long delay means a long wait for the number. Treat the count as a record of completed exits rather than a live view of how many people have been suppressed.
Second, time-based exit filters are evaluated at the end of the delay. In the documentation's example, the exit criteria are checked at the conclusion of the Delay step's duration, so a user who orders soon after entering a 24 hour delay can fail a "placed an order less than 1 hour ago" test. For long delays, prefer an event-based exit over a recency window.
Exception events include placing an order, starting a session, performing a custom event and performing a conversion event. You can also add segments and filters, and users who match them exit and receive no further messaging.
When should you use Action Paths instead of exit criteria?
Exit criteria end a journey. Action Paths sort users onto different branches and keep them in the Canvas. The table sets the two side by side.
| Question | Exit criteria | Action Paths |
|---|---|---|
| What happens to the user? | Leaves the Canvas | Moves down a branch and stays in the Canvas |
| When does it act? | The exit event marks the user at once, but a user in a Delay step leaves when the delay ends | The first qualifying action sends the user down that group, unless ranking is on |
| If the user does nothing | Nothing changes and the user keeps moving | The user takes the Everyone Else group when the evaluation window ends |
| Longest hold | A Delay step can run up to 730 days | The evaluation window can run up to 31 days |
Action Paths evaluate for one day by default, and you can adjust the window by seconds, minutes, hours, days and weeks. Without ranking, a user who performs the trigger event for any action group advances through the first qualifying group. With ranking on, everyone is held to the end of the window and takes the highest priority group they qualify for.
Braze documents a pattern for matching an exit to the entry event. In an abandoned ticket example, a user who purchases the same ticket they originally selected exits and stops receiving reminders, while a user who buys a different ticket stays in the Canvas and continues to receive messages about the original one. The same comparison works in Action Paths, which lets a different purchase take a different branch.
What does a hypothetical retention journey look like?
Take a hypothetical meal-kit brand whose customers can pause deliveries. The pause is a custom event, which makes it a clean action-based entry.
The example below is illustrative only. It uses no client data and carries no figures, because the documentation gives none for retention timing.
- Entry: action-based, on the pause event. Entry controls allow re-entry, since a customer can pause more than once.
- Exit criteria: the customer resumes deliveries. The resume event's plan property is compared with the plan property from the pause event, using the pattern for matching exit criteria to entry events.
- Delay: a wait chosen by the brand. The documentation sets no right length, so the brand tests it.
- Message: a reminder about the paused plan.
- Action Paths: a window of a few days with three groups, one for customers who resume a different plan, one for customers who start a session, and Everyone Else.
- Final message: the Everyone Else group gets a note that asks whether they want to keep the pause or end the plan.
The shape matters more than the detail. Exit and branching decisions sit before every message, so each message reaches only people the journey hasn't already resolved.
How do you know reducing churn with Braze is working?
Add a control group. Braze tracks the conversions for users placed into the control group, although they don't receive any messages. That gives you a comparison that a before-and-after reading can't.
Conversion windows need the same planning as delays. Braze recommends setting the conversion window to at least match the longest path in the Canvas. Conversions are tracked even after the Canvas ends, while exits are not tracked once the user has exited, so the two figures can diverge when the exit event and the conversion event are the same.
Exits and conversions are tracked on different terms, so read the two figures side by side. If you want help judging whether a retention Canvas is working, CustomerIK's marketing analytics service covers churn prediction and lifetime value forecasting.
What tends to go wrong in retention Canvases?
Four failure modes recur, and each has a documented cause.
The first is the pause that loses people. When you stop a Canvas, users already waiting in a Delay step are not exited immediately.
If the delay window has passed while the Canvas was stopped, those users exit instead of receiving the next step. Check where your users are waiting, and how long the longest delay runs, before you stop a live journey.
The second is the early clock. If an action-based Canvas sends sooner than expected, check that the custom event timestamp carries the current time rather than a backdated one. Braze evaluates the delay from the timestamp sent with the event.
The third is the self-referencing trigger described earlier, where the entry event is also an audience filter.
The fourth is the ranking flip. Braze's guidance is not to change the ranked order setting after launch when users are already in the step. If you need different ranking, build a new Action Paths step.
How does this connect to the rest of the programme?
Retention automation sits on top of clean event data, so tracking user events for better campaigns comes first. Lapse and pause signals are only as good as the events behind them. Once a journey exists, setting up engagement triggers covers the entry side in more detail.
When a customer has already gone quiet, the win-back campaigns approach picks up where a retention Canvas leaves off. Optimising push notification timing matters for any journey that uses push as one of its steps. For a wider view, read top customer retention strategies using Braze and automating loyalty programmes with points, tiers and offers.
A write-up of re-engaging lapsing users with an interactive reward covers a lapsing-user programme built into automated lifecycle paths. It is a delivery view of the same lapse-and-return pattern.
The mechanics in this post come from the Braze documentation on exit criteria. Limits and defaults can change, so check the current page before you build.
What should you build first?
As a Braze implementation partner, CustomerIK recommends building the exit and the measurement before the first message. Write down the churn definition, choose the exit event and the conversion window, add a control group, and only then draft the messages.
Frequently Asked Questions
1. How long can a Delay step in a retention Canvas run?
Delay steps can run up to two years (730 days). An Action Paths evaluation window is shorter, with a maximum of 31 days.
2. Does a user who meets the exit criteria stop receiving messages straight away?
Braze marks them to exit immediately and they don't advance to later steps. If they're in a Delay step they remain there until the delay ends, then they leave.
3. Can exit criteria use segments and filters?
Users who match the segment and filter exit the Canvas and don't receive any further messaging.
4. What happens to users waiting in a delay if I stop the Canvas?
They aren't exited immediately. If the delay window passes while the Canvas is stopped, those users exit instead of receiving the next step.
5. Should a retention Canvas have a control group?
Braze tracks conversions for control group users, who receive no messages. That gives a like-for-like comparison with the users who were messaged.
Sources
- Braze, Exit criteria. Retrieved 21 Oct 2026.
- Braze, Delay. Retrieved 21 Oct 2026.
- Braze, Action Paths. Retrieved 21 Oct 2026.
- Braze, Create a Canvas. Retrieved 21 Oct 2026.
- Braze, Matching exit criteria to entry events. Retrieved 21 Oct 2026.
- Braze, Canvas best practices. Retrieved 21 Oct 2026.
- Limits and defaults on these pages can change, so check each page before you build.
- Written by CustomerIK, a Braze implementation partner.
CustomerIK is a Braze implementation partner working across onboarding, technical integration, marketing operations and customer data management.
If you are mapping a retention Canvas and want a second view on the exit and measurement design, let's talk.
.svg)
