Before this engagement, the streaming platform was facing a genuinely urgent set of interconnected problems:
Product and marketing leadership could see the churn problem clearly in their subscriber numbers, but had no systematic way to intervene before a subscriber actually canceled -- by the time churn was visible in reporting, it was already too late to act on that specific subscriber. The team had a general sense that onboarding mattered, but no structured process for acting on that intuition, and messaging decisions were largely made campaign by campaign rather than as part of a coherent lifecycle strategy.
We designed and implemented a lifecycle-driven Braze engagement program built around genuine retention signals rather than blanket messaging, addressing onboarding, risk identification, and channel discipline as one coordinated system:
Each of these four workstreams targeted a different point in the subscriber lifecycle -- onboarding, risk identification, channel discipline, and preference management -- but were designed to work together as one coherent system rather than four separate initiatives. The churn risk scoring, for instance, directly fed which subscribers entered the re-engagement Canvas, and the frequency capping logic applied consistently across both the onboarding and re-engagement journeys to prevent either from over-messaging a subscriber already receiving other communications.
Building the churn-risk scoring model required genuine patience before it could be trusted operationally. Rather than launching intervention Canvases against a scoring model built on assumptions, the team first ran the scoring logic in a monitoring-only mode for several weeks, comparing its predictions against actual subsequent cancellations before activating any customer-facing intervention based on it. This validation period meant the eventual re-engagement Canvas was acting on a model with demonstrated predictive accuracy, not an untested hypothesis about what churn risk actually looked like for this specific subscriber base.
Within five months of full implementation, once the churn-scoring model had enough historical data to identify genuine risk patterns and the onboarding and re-engagement Canvases had matured through their initial testing period:
Analysis of historical subscriber data revealed a pattern the platform hadn't previously acted on systematically: subscribers who watched meaningful content within their first 48 hours after signup had substantially lower churn rates over their first three months than those who didn't. This made the onboarding Canvas's decision-split logic genuinely consequential -- rather than treating all new subscribers identically, the journey specifically detected whether that critical early engagement had happened, and responded with more targeted, higher-effort re-engagement messaging for subscribers who hadn't yet started watching. This early intervention, timed to the specific window where it mattered most, was directly responsible for a meaningful share of the overall churn reduction, and gave the team a genuinely evidence-based rationale for where to focus limited engineering and content resources going forward.
A common assumption in subscription businesses is that discount incentives are the most effective churn-prevention lever. Rather than assuming this, the team explicitly tested personalized content nudges against direct incentive offers within the re-engagement Canvas, using a genuine A/B structure rather than committing to one approach based on intuition. The results were genuinely instructive: for subscribers whose declining engagement stemmed from not finding content that matched their interests, personalized recommendations outperformed discount offers -- while for subscribers citing cost sensitivity in exit surveys, incentives performed better. This led to a more sophisticated, segmented re-engagement approach rather than a single one-size-fits-all intervention, directly informed by what the data actually showed rather than an untested assumption.
A genuine risk in running both an onboarding Canvas and a separate churn-risk re-engagement Canvas simultaneously is that a subscriber could theoretically qualify for both at once -- a new subscriber who signs up, immediately shows low initial engagement, and gets caught by both journeys' entry criteria within the same short window. Without deliberate coordination, this could mean a genuinely overwhelming volume of messages arriving in quick succession, undermining the disciplined channel approach the rest of the program was built around. The team addressed this by applying a shared frequency cap across both Canvases rather than managing each independently, and by building explicit exclusion logic so a subscriber actively in the onboarding journey wouldn't simultaneously enter the churn re-engagement flow. This kind of cross-Canvas coordination is easy to overlook when journeys are built and tested in isolation, but became a genuine focus once multiple concurrent journeys were live simultaneously.
Once the churn-scoring model proved reliable, the team found a genuinely valuable secondary application beyond retention messaging specifically: using the same engagement signals to evaluate whether the platform's content recommendation engine was actually serving relevant suggestions to different subscriber segments. Subscribers with declining engagement despite active recommendation delivery pointed to a recommendation quality gap, not just a general disengagement pattern -- a distinction the messaging team could only make once they had reliable engagement scoring to compare against. This finding was shared with the product team responsible for the recommendation engine itself, becoming an unplanned but genuinely valuable cross-team insight that extended beyond the original messaging engagement scope.
An unexpected benefit of introducing genre-specific Subscription Groups was the aggregate preference data it generated -- for the first time, the platform had a genuinely reliable, opt-in-based signal of which content categories subscribers actively wanted to hear about, distinct from passive viewing history alone. This data proved valuable beyond messaging targeting: content licensing and acquisition discussions began referencing genre subscription rates as one input alongside viewing metrics, since active opt-in arguably reflects a more deliberate signal of subscriber interest than viewing behavior that might simply reflect whatever content happened to be available at any given time. This wasn't the original purpose of the Subscription Groups rollout, but became a genuinely valued secondary data source once the groups had accumulated meaningful subscriber participation over several months.
Taken together, the churn reduction achieved here reflects less a single clever tactic and more a genuine shift in how the platform approached subscriber communication -- from reactive, undifferentiated messaging toward a coordinated system that respected both subscriber preference and the specific moments where intervention could actually change an outcome. That shift in underlying approach, more than any individual Canvas or flow, is what leadership credited as the real driver of sustained improvement across the whole subscriber base.
Our team of experts is here to help you achieve measurable, lasting results.