Retention Metrics Every Marketer Should Track in Braze

The Braze retention metrics worth tracking are rolling retention, range retention and retention against a control group, all built on one retention event that reflects real value. Read them knowing how Braze counts. Rolling retention credits any return on or after a day, range retention credits returns within a window, and the summary row can count a user twice.

Key Takeaways

• Track rolling retention, range retention and retention against a control, all on one retention event that reflects real value.
• Braze retention reports measure the retention event over the 30 days after a user receives the campaign or Canvas.
• Rolling retention counts returns on or after each day; range retention counts returns within each window.
• The All Users row counts a user twice if they received the message in more than one cohort.
• A report can't be re-run with the same retention event for 24 hours, and isn't available for API-triggered campaigns.
• Read retention by variant against the control group, with its confidence figure.

Which retention metrics should you track in Braze?

Track a small set, and know what each one answers. Braze's retention report sits on the analytics page of each campaign or Canvas, and it measures whether recipients performed your chosen retention event over the 30 days after receiving the message.

Our post on customer retention strategies covers what to do with these numbers across a programme. This post covers how to read them without misleading yourself.

How do rolling and range retention differ?

They count the same user differently. Rolling retention counts users who did the retention event on or after each day shown, so a user who returned between days three and seven counts in the day three, day one and day zero columns.

Range retention counts users who returned within each window. A user who returned between days three and seven, then again on day 13, counts in both the Day 3-7 and Day 7-14 ranges.

Use rolling retention to answer whether users are still around after a point in time. Use range retention to see when users come back, which is the better view for spotting where a journey's effect fades.

How should you choose the retention event?

The retention event decides what the report measures. Choose the action that shows a customer is getting value, rather than the easiest event to log.

A session start is simple, but a user can open the app and leave. A purchase, a key feature use or a content play usually says more about whether retention is real.

You can switch the retention event and re-run the report to look at different outcomes. Agree one primary retention event per programme, though, so results stay comparable over time.

What can make a retention report misleading?

Several documented behaviours change how the numbers read. The All Users row counts a user twice if they received the campaign or Canvas in more than one cohort, so it can overstate reach for recurring sends.

Incomplete values are marked with an asterisk, because recent cohorts haven't had the full window yet. The percentages are weighted averages, and the report only lists days on which the campaign or Canvas was sending messages.

The date range includes users who entered during that window, then shows those who performed the retention event within it. Retention reports also aren't available for API-triggered campaigns.

Two practical limits shape your workflow. A report can take up to five minutes to load, and you can't re-run it with the same retention event for 24 hours.

How do you prove a message improved retention?

Compare against the control group. The by-variant view shows rolling retention for each variant and the control, with the percentage change from control and a confidence figure.

Braze calculates that confidence by comparing each variant's rate with the control's using a Z test. Treat a large change with low confidence as a reason to keep testing, not a result.

The control group's own retention is worth reading by itself. It shows what retention looks like without the message, which is the baseline every retention claim should start from.

What should you look for once the report is running?

Braze suggests several questions for interpreting retention reports. Each points to a different action.

1. Do cohorts from different days of the week perform differently for recurring campaigns?
2. Where does the impact start to decline, and should a new message target that point?
3. Is there a sign of messaging fatigue?
4. Did a change you made a few days ago move retention?

The second question is the most useful for journey design. The day retention starts to fall is often the right moment for the next step in a Canvas.

In what order should you set up retention measurement?

Set it up in this order so the comparison exists when you need it.

1. Choose one primary retention event that reflects real value.
2. Add a control group to the campaign or Canvas before launch.
3. Decide whether the question needs rolling or range retention.
4. Run the report once cohorts have had time to complete, and ignore values marked with an asterisk.
5. Read retention by variant against the control, with its confidence.
6. Note the day retention declines, and test a message at that point.

Which mistakes should you avoid?

Name the mistakes so they're easy to spot. A shallow event measures retention on an action that doesn't show value, so the figure looks healthy while customers drift.

A double count reads the All Users row as unique people for a recurring send. An early read judges cohorts still marked as incomplete.

A missing baseline reports retention without the control group, so nobody can tell whether the message made a difference.

Where does this connect to the rest of the programme?

Retention reports measure the journeys covered in building effective win-back campaigns. Retention events only measure well when they're logged well, as set out in tracking user events in Braze.

Comparing variants is covered in A/B testing your email campaigns, and email engagement figures in email campaign reporting.

As a Braze implementation partner, CustomerIK recommends choosing the retention event before the report. The right event and a control group turn a retention report into evidence.

A retention report with a fixed cohort

Use a fictional cohort of 1,000 users who completed setup on the same day. Choose one meaningful return event and one observation window before reading the report. Keep rolling and range retention separate; they answer different questions.

If 240 of 1,000 treatment users and 220 of 1,000 control users meet the same retention definition, the observed rates are 24% and 22%. That is a two-percentage-point difference, not proof of significance. Check allocation, uncertainty and equal follow-up before claiming incremental lift.

The worked example and test cases are illustrative editorial guidance, not measured client outcomes or a claim that a live integration has been tested.

Frequently Asked Questions

1. What is the difference between rolling and range retention in Braze?

Rolling retention counts users who returned on or after each day. Range retention counts users who returned within each window, so one user can appear in several ranges.

2. How far does a Braze retention report look?

It measures whether recipients performed the retention event over the 30 days after receiving the campaign or Canvas.

3. Why does the All Users row look higher than expected?

A user who received the campaign or Canvas in more than one cohort is counted twice in that row.

4. Can I re-run a retention report straight away?

Not with the same retention event for 24 hours. You can change the retention event and run it again.

5. Are retention reports available for API-triggered campaigns?

They aren't available for API-triggered campaigns.

Sources

• Braze, Retention reports. Retrieved 18 Oct 2026.

CustomerIK is a Braze implementation partner working across onboarding, technical integration, marketing operations and customer data management.

If you're setting up retention measurement and want the event and control group right from the start, let's talk.