How RTL+ used BrazeAI Agent Console™ to personalize streaming recommendations at scale

RTL+ is one of Germany’s most popular streaming services, offering content across reality television, live sports—including the UEFA Europa League, Premier League, and NFL—and original entertainment content that includes crime, comedy, and action programming. With approximately 7 million subscribers on big screen, web, and mobile, RTL+ operates in a highly competitive streaming landscape where engagement and retention are defining metrics. The platform offers three subscription tiers and runs a cancel-to-keep model in which subscribers must actively cancel within three days of purchase.

The goal is to keep viewers on the site, discovering new content. Sometimes this is easy. For example, strong segmentation makes it easy to serve viewers of the popular Bachelor or Bachelor in Paradise franchise similar content within the genre, but other viewers log in specifically to watch a single sports match and then disappear. “This makes onboarding very important for us,” says Lea Engelmann, who leads lifecycle marketing at RTL+.

Strong segmentation with limited personalization

RTL+ previously built sophisticated segmentation within Braze. The lifecycle marketing team used custom attributes and behavioral events to ensure campaigns reached the right audiences, triggering in-app messages when a user clicked cancel, for example, or targeting genre-specific content to subscribers with matching attributes. The segmentation helped move the needle on engagement and churn prevention, but the team knew there was still room for improvement.

They also knew that having well-defined audiences didn’t mean that the messaging was resonating. The actual content of each campaign remained largely uniform, addressing each broad segment. This meant that a loyal fan of crime drama and a casual viewer of reality shows might receive an email that looked nearly identical. “We reach the right users, with the right trigger, at the right time, but within the message, the messaging remains generic,” Engelmann says. Product-driven recommendation logic could surface “continue watching” modules, but the copy, tone, and framing of the message didn’t adapt to the individual. That’s what RTL+ sought to change by building an agent within BrazeAI Agent Console™.

Applying agent personalization across 3 key journeys

Engelmann prioritized three journeys for AI-powered personalization: onboarding, churn prevention, and win-back campaigns. Each represents a high stakes moment for the business, when the right message can make a difference in whether the viewer finds value in the streaming service and its content.

To power personalization, Engelmann built an agent in BrazeAI Agent Console™ that draws on approximately 20 customer data points per user. These signals include top genre affinity (both the single strongest genre and the top five), recent usage intensity (usage levels tracked over seven-day and thirty-day windows), format diversity (the number of content types a subscriber streams in parallel), first-video-view genre, and customer-declared interest data captured during onboarding. Sports-specific affinity signals—such as soccer or NFL preference—are also passed to the agent for relevant audiences.

The agent generated personalized message components that were dynamically assembled into RTL+’s existing Braze templates using Liquid, custom HTML, and custom code layouts.

The goal of the agent was to evaluate a broad set of customer attributes and behavioral signals and then adapt the messaging to each individual subscriber. Outputs included personalized subject lines, preheaders, headlines, recommendation framing, editorial copy, and content prioritization within the email.

The agent was provided with approximately 20 customer signals, including genre affinities, sports interests, usage intensity, content diversity, first-viewed genre, onboarding preferences, subscription history, and other behavioral attributes stored in Braze.

For example, rather than showing every subscriber the same welcome message, the agent could recognize that a customer had subscribed before and generate copy such as “Welcome back, Lea—great to have you with us again.” It could also identify a subscriber's strongest genre affinities and adapt both the messaging and the content presentation accordingly. A reality fan might receive messaging focused on Reality content, while Romance and True Crime recommendations would be prioritized further down the email based on additional affinities. If Sports viewing behavior was detected, sports-related content modules or banners could also be surfaced.

The agent also adapted its tone and framing based on engagement levels. For lower-engagement users, messaging became more activation-focused, while highly engaged users received copy that acknowledged their existing interests (for example, “You're clearly a big Reality fan…”).

The agent also uses a signal hierarchy: Strong, recent behavioral data takes precedence. But where data is sparse, the agent falls back to softer signals or brand-appropriate defaults. This ensures that new subscribers with limited behavioral history still receive warm, relevant messaging.

In short, the agent's role was to transform customer data into individualized messaging and content prioritization at scale, enabling a level of personalization that would not have been feasible through manually created campaign variants.

The initial prompt architecture was developed using an enterprise ChatGPT instance before switching to BrazeAI Operator™ and augmenting with brand voice and tone guidelines and messaging frameworks to further improve output quality and reduce manual prompt maintenance.

Liquid templating within Braze assembles the agent’s copy outputs into fully rendered emails and messages while custom HTML handles in-app message formats. Together, each campaign generates individually tailored content at send time, replacing what would otherwise require dozens of manually created variants. For a high-volume platform like RTL+, there are gains in both operational efficiency and quality improvement, thanks to agent personalization.

The results: 21% higher watch engagement and 40% higher open rates in RTL+’s low-engagement journey

RTL+ first deployed the agent within its low-engagement lifecycle journey, targeting subscribers whose streaming activity had significantly declined. Compared with a no-message campaign control group, users exposed to the agent-powered journey generated approximately 21% higher watch engagement across the measured period.

The campaign also generated a 40% uplift in open rates, demonstrating that agent-powered personalization increased the perceived relevance of the messaging and encouraged more subscribers to engage with RTL+ content.

Even so, the direction is consistent: agent-personalized messaging measurably increases content consumption. And for a team already invested in segmentation quality, the uplift confirms that closing the personalization gap at the messaging level makes a difference.

RTL+ is continuing to test and refine personalization across onboarding journeys. Early learnings suggest that personalization becomes increasingly effective as more behavioral signals become available, helping guide the next phase of development.

Now, Engelmann’s team at RTL+ is building their second iteration of agent personalization, introducing content-level metadata over genre-level signals. The agent is fed information about specific formats and episodes—along with the platform’s recommendation engine outputs—to deliver messaging that reads less like “We see you like reality shows. Explore our newest ones” and more “A new episode of The Bachelor with Tim and Sebastian is coming up this week. One of them is already falling in love.” The team anticipates that this shift from genre to specific content recommendations will drive further engagement.

“With agent-powered personalization, we're making our messaging far more relevant for every individual subscriber. The Braze agent helps us close the gap between strong audience segmentation and truly personalized communication at scale.”

Lea Engelmann

Topic Lead Engagement, RTL+

Key takeaways

1. Strong segmentation and message personalization are complementary: RTL+ had world-class segmenting already in place, yet still saw significant engagement uplift after closing the personalization gap at the messaging level.
2. Both signal hierarchy and volume matter: With approximately 20 customer data points in play, RTL+ built a fallback architecture that ensures every user receives a relevant message, regardless of how much behavioral data they have.
3. One campaign, with infinite variants: By combining BrazeAI Agent Console™ with Liquid templating, RTL+ now easily generates individually tailored messages within a single campaign, which was never possible before now.
4. The next phase is the move towards content specificity: The move from genre-level to content-level personalization is game-changing, allowing RTL+ to recommend specific shows and upcoming episodes that really capture a viewer’s attention.

FEATURED TOPICS