Use case: Predict subscription upgrades with smarter targeting

This example shows how a fictional brand uses Braze Predictive Events to define the outcomes that matter to their business—like upgrading to a pro membership—and build targeted strategies that improve results.

Let’s say Jordan is a lifecycle strategist at Steppington, a health and fitness app with free and paid tiers. Jordan’s team has a goal to increase Pro plan upgrades without bombarding their entire free user base with discount messages. Currently, they send a “Try Pro for 50% off” promo to every free-tier user after seven days. While it drives some conversions (about 5% over 7 days), it also results in excessive outreach—including discounting users who were likely to upgrade anyway.

To improve targeting and reduce messaging fatigue, Jordan uses Predictive Events to model the likelihood that a user will upgrade to Pro within the next 7 days. He defines a custom event: upgraded_to_pro, then uses that to train a prediction model and segment users into smart, action-oriented groups.

This tutorial walks through how Jordan created:

• A predictive model for upgraded_to_pro within 7 days
• Segments that help increase conversions while sending fewer total messages

Step 1: Create a predictive model for upgrades

Jordan starts by defining the outcome that matters most for his upgrade strategy: a user moving from the free tier to Pro. Rather than relying on generic triggers like “time since signup,” he wants to forecast which users are actually likely to convert. This way, his team can act on real signals, not just assumptions.

1. In the Braze dashboard, Jordan goes to Analytics > Predictive Events.
2. He creates a new event prediction and names it “Upgrade to Pro in 7 days”
3. As the target event, he selects his custom event: upgraded_to_pro.
4. Jordan sets the prediction window to 7 days, sets an update schedule, and creates the prediction.

Step 2: Segment users based on upgrade probability

After training is complete, Braze assigns an Event Likelihood Score (0–100) to each eligible user. Jordan uses this score to create actionable segments—one for high-intent users who may not need a discount, and another for users who likely won’t convert without support.

1. Jordan navigates to Segments in Braze.
2. He creates two segments using the Event Likelihood Score filter and selects the prediction he created. The two segments are:
   • Likely to upgrade: Score more than 70
   • Needs nudge to upgrade: Score more than 40 and less than 70

Step 3: Personalize messaging by intent level

Now that Jordan has clear upgrade intent signals—and refined subgroups based on user behavior—he builds a messaging strategy that adapts to what each user needs. No more one-size-fits-all blasts.

He chooses email as the primary channel for this campaign. Why? Because Jordan wants to explain the value of Pro for high-intent users and make a compelling case to more hesitant users—both of which require space, visuals, and a strong CTA. Email gives him the flexibility to do that well without pressuring users, and lets him track performance through click behavior.

Jordan creates a Canvas that splits the experience based on the segments he just built. He adds an Audience Paths step to target:

• High intent, fitness-focused users
• High intent, other users
• Low intent, fitness-focused users
• Low intent, other users

He also sets the Canvas conversion event to the custom event upgraded_to_pro, so Braze tracks upgrade conversions automatically as users progress through the flow.

Example messages per path

• high intent, fitness
• high intent, other
• low intent, fitness
• low intent, other

Step 4: Measure results and optimize your strategy

After the campaign runs, Jordan reviews performance in Canvas Analytics to understand how well the personalized paths performed—and whether combining predictive intent with behavioral signals improved upgrade rates.

Email performance by path:

• High intent, fitness
  • Open Rate: 34%
  • Click Rate: 20%
  • Conversion Rate: 13%
  • No discount used
• High intent, other
  • Open Rate: 30%
  • Click Rate: 17%
  • Conversion Rate: 11%
  • No discount used
• Low intent, fitness
  • Open Rate: 27%
  • Click Rate: 12%
  • Conversion Rate: 8%
  • 50% off offer included
• Low intent, other
  • Open Rate: 23%
  • Click Rate: 9%
  • Conversion Rate: 6%
  • 50% off offer included

Compared to the team’s previous one-size-fits-all campaign (where a blanket discount after 7 days led to just 5% conversions and over-messaging) the targeted approach shows meaningful lift across all groups, with improved efficiency and fewer unnecessary discounts.

The funnel report also shows a clear reduction in drop-off across key steps, particularly for low-intent users who received personalized messaging. More users are opening, clicking, and upgrading—proving the value of intent-based targeting.

Jordan uses these insights to:

• Explore A/B tests on subject lines and CTA phrasing
• Reassess the discount threshold for mid-intent users
• Continue refining segments based on additional behaviors like content views or app feature usage

With Predictive Events and layered segmentation, his team now has a scalable strategy that adapts messaging based on user intent and behavior—driving more upgrades while preserving brand trust.