Understanding How the AI Audiences' Models Power Your Campaigns

Our AI models for Wunderkind Text help you send smarter, more personalized messages to your customers — automatically. This article explains how AI Audiences (Purchase Propensity and Product Affinity) and Send-Time Optimization work, what data they use, and how they differ from traditional segmentation.

Purchase Propensity

What it does

Predicts how likely a person is to make a purchase soon. This helps you prioritize who receives conversion-focused campaigns.

How it works

• Aggregates recent user behaviors, including:
  • Add to Cart events
  • View Item and View Category events
  • Site visits
  • Past purchases
• Only analyzes the last 14 days of activity (older data has less predictive power).
• Uses machine learning trained on aggregated, anonymized data from customers who have opted into the shared graph.
• Assigns each Privacy ID (an anonymized user) a likelihood score of purchasing.

Why it’s powerful

Unlike rule-based segments (for example, “opened 3 emails + clicked 2 products”), the model identifies hidden behavioral patterns that historically lead to conversion. Scores update automatically as new data comes in — meaning your targeting always reflects the most recent intent signals.

Product Affinity

What it does

Measures how strongly a person is interested in a product or related products.
Use it to personalize recommendations and product-focused campaigns.

How it works

• Looks at users with at least one of the following:
  • Add to Cart event
  • Conversion
  • View Item event
• Uses a 120-day look back window to capture longer-term engagement.
• Is a generalized model, not specific to one client — trained using aggregated data across all customers on the shared graph.
• Combines multiple techniques:
  • Collaborative filtering to connect users and related products
  • Natural language and image models to find product relationships

No data is shared externally. All data remains anonymized and is not shared with Google or any third parties.

Why it’s powerful

Product Affinity goes beyond browsing history — it learns from large-scale behavior patterns to recommend what each shopper is most likely to care about next.

Send-Time Optimization

What it does

Predicts the best time and channel (email or text) to send a message for maximum conversions.

How it works

• Analyzes multiple activity types:
  • Page views, Add to Cart, and conversion events
  • Message-level engagement (opens, clicks, and whether they led to conversions)
• Optimizes for conversions, not just clicks — based on updated testing results.
• Recalculates regularly to adapt to user behavior over time.

Channel logic

Our updated channel logic ensures messages reach customers through the most effective channel — without missing key opportunities.

• If a user is most likely to convert on text, they will now receive both a text and an email.

  • This change was made after observing that only sending texts caused a noticeable drop in email engagement and revenue.

• If a user is most likely to convert on email, they will only receive the email.

  • Sending emails has no additional cost impact for clients, unlike SMS.

This approach balances performance and cost efficiency, ensuring high-performing channels are prioritized while maintaining overall reach.

Why it’s powerful

Instead of guessing “what time works best,” the model personalizes send times for each recipient — helping your messages reach people when they’re most likely to take action.

How these models work together

Together, these models help you send the right message, about the right product, at the right time — automatically.

Data privacy and security

We take privacy seriously.
All models are built with privacy-first principles:

• Uses Privacy IDs instead of identifiable user data
• No emails, phone numbers, or behavioral events are ever shared between clients
• Aggregated data may inform generalized algorithms, but individual data remains isolated
• Data is securely processed and stored using standard cloud infrastructure

Learn more

For how-to instructions, visit Text Message Segmentation: Purchase Propensity and Product Affinity User Guide. Or contact your Customer Success Manager to learn how to get the most from these models.