How to Predict Customer Churn with SFMC+Einstein

Customer churn is a reality every industry contends with, but smart businesses don’t just accept it, they fight it head-on.

Staying ahead means keeping a close watch on how others are retaining their customers, and more importantly, learning how to predict customer churn accurately.

That’s where predictive tools like Einstein AI in SFMC come in. At Mavlers, we’ve used Einstein AI in over 800 SFMC client accounts for a variety of purposes, including churn prediction in Salesforce Marketing Cloud.

With the right insights, you can launch timely re-engagement campaigns that make a real impact. Let’s find out how.

Churn prediction in salesforce marketing cloud

Predict churn risk with sfmc einstein

Triggering einstein-powered re-engagement

Einstein ai with re-engagement journeys in sfmc

Wrapping Up

Churn prediction in Salesforce Marketing Cloud

To build an effective churn prediction system, three elements are required to set up the foundation:

1. Quality Data: Comprehensive historical data which includes customer interactions, purchase history, service engagements, and behavioral patterns.

2. Analytical Acumen: The ability to prepare and interpret data, whether through in-house expertise or external partners.

3. Enabling Technology: A platform capable of handling large datasets, running complex machine learning algorithms, and operationalizing predictions.

Source: Customer Gauge

Salesforce empowers businesses to leverage these advanced capabilities. That’s why churn prediction in Salesforce Marketing Cloud has become more accessible than ever.

Predict churn risk with SFMC Einstein

Einstein Discovery

It is Einstein Discovery that provides the analytical engine for understanding and predicting churn. 

Einstein Discovery is a part of the Salesforce Einstein AI suite that uses machine learning to analyze large datasets, uncover hidden patterns, and provide predictive insights. 

As far as churn prediction is concerned, the tool can: 

• Analyze customer behavior: It ingests vast amounts of data from your SFMC instance, including email engagement, website interactions, purchase history, or support case data.
  
• Identify churn indicators: It learns from historical data of customers who have churned and identifies the specific factors and patterns that most strongly correlate with churn risk.
  
• Predict churn likelihood: Based on these patterns, Einstein Discovery can then predict the probability of individual customers churning in the future.
  
• Provide actionable insights: Beyond a score, it can explain why a customer is at risk of churning, highlighting the top contributing factors. This allows marketers to understand the root causes and develop targeted retention strategies.

To churn risk with SFMC Einstein using Einstein Discovery, you need to build a churn prediction model first. 

Building a churn prediction model with Einstein Discovery

The journey to building a powerful churn prediction model with Einstein Discovery involves:

1. Data Preparation: This often constitutes the lion’s share of the effort. It involves meticulous collection, cleaning, and transformation of data from all relevant sources into a comprehensive dataset. Addressing missing values, deduplication, and enriching data are vital steps to ensure the model learns from accurate and complete information.

2. Story Tuning: Einstein Discovery generates an interactive “story” that outlines the model’s findings, highlighting key drivers of churn in an intuitive, natural language format. This iterative phase involves refining the model by evaluating its insights, assessing statistical significance, and addressing potential issues like overfitting (where the model is too tailored to historical data and performs poorly on new data).

3. Model Evaluation: Rigorous evaluation ensures the model’s reliability. Key metrics include the following:

• Confusion Matrix:This is a table showing true positives (correctly identified churners), true negatives (correctly identified non-churners), false positives (predicted churn, but no churn), and false negatives (predicted no churn, but churn occurred). The goal is often to minimize false negatives in churn prediction, as missing an at-risk customer is more costly.

• ROC Curve and AUC (Area Under the Curve): This measures the model’s ability to discriminate between churners and non-churners. An AUC value closer to 1 signifies a highly accurate model.

• Gains Chart/Lift Chart: Illustrates the model’s efficiency in identifying at-risk customers compared to random selection.

It’s critical to understand that a predictive model is never a “set it and forget it” solution. Continuous monitoring and recalibration are essential to adapt to data drift – the natural evolution of customer behavior and market dynamics over time.

Triggering Einstein-powered re-engagement

Once Einstein Discovery has assigned a churn risk score to customers, the next step is to convert this intelligence into proactive customer re-engagement.

1. Operationalizing the predictions

Churn risk scores are seamlessly embedded within Salesforce CRM. They can appear directly on Lightning Pages associated with customer records, in list views for quick segmentation, or within reports for broader analysis. 

Beyond just a score, Einstein Discovery components can also display the top contributing factors to a customer’s churn risk and even suggest prescriptive “next best actions”. 

While Einstein Discovery provides a holistic churn probability based on all available customer data, Einstein Engagement Scoring operates directly within Marketing Cloud to provide granular, marketing-specific insights. It analyzes historical email and mobile/push interactions to predict a subscriber’s likelihood to open, click, or even unsubscribe from messages. 

These insights are essential when you want to predict and re-engage inactive users through intelligent campaign strategies. Combining Einstein Engagement Scoring for re-engagement with overall churn prediction allows for a smarter retention approach.

2. Synchronizing data with SFMC

The churn risk scores, along with relevant customer data from Salesforce CRM, are automatically synchronized with SFMC. This is achieved through robust, out-of-the-box connectors or via API integrations, populating dedicated Data Extensions within SFMC. 

This allows you to trigger re-engagement emails Salesforce teams can use across touchpoints for high-risk customers. Having this synchronized data fuels Salesforce Marketing Cloud re-engagement automation that adapts in real-time to changing customer behaviors.

Source: Trailhead

3. Orchestrating journeys in SFMC journey builder

SFMC Journey Builder for churned users becomes incredibly powerful when enhanced with churn risk data.

In re-engagement journeys in SFMC, you can:

• Trigger Events: A customer’s churn risk score crossing a predefined threshold (e.g., from “low” to “medium” or “high”) can automatically enroll them into a specific re-engagement journey.
  
• Decision Splits: Journeys can incorporate decision splits to tailor paths based on the exact churn score, the underlying reasons for churn identified by Einstein Discovery (e.g., service issues vs. product dissatisfaction), or the customer’s historical value.
  
• Personalized Content: Using personalization capabilities, messages can be dynamically tailored. If Einstein identifies “lack of recent engagement” as a key churn factor, the re-engagement email might highlight new features, relevant content, or exclusive access to community events.
  
• Multi-Channel Engagement: You can deploy a sequence of coordinated touchpoints across various channels, such as email, in-app messaging, and SMS. 

This approach helps brands predict and re-engage inactive users with greater precision and timeliness. 

Einstein Engagement Scoring can also be embedded in these journeys to further refine outreach based on recent trends.

4. Einstein engagement frequency & send time optimization

SFMC’s automation also leverages other Einstein capabilities. 

• Einstein Engagement Frequency helps determine the optimal number of messages to send to each individual to prevent saturation and disengagement. 

Source: Salesforce Ben

• Einstein Send Time Optimization predicts the best time to send messages to maximize open and click rates for each subscriber, enhancing the effectiveness of re-engagement efforts within the journey.

Consolidating Einstein Ai with re-engagement journeys in SFMC

The combination of Einstein’s predictive intelligence and SFMC’s journey orchestration capabilities delivers a powerful, proactive customer retention framework:

• Proactive Churn Prevention: Reach out to at-risk customers before they decide to leave, shifting from reactive to preventive customer management.

• Hyper-Personalized Re-Engagement: Deliver truly relevant messages and offers based on the specific reasons a customer is likely to churn, increasing the effectiveness of re-engagement efforts.

• Enhanced Customer Lifetime Value: By retaining more customers, businesses naturally boost their CLTV and overall revenue.

• Optimized Resource Allocation: Focus sales, service, and marketing efforts on the most critical customer segments, ensuring maximum ROI from retention initiatives.

• Unified Customer View: Break down departmental silos by ensuring all customer-facing teams operate from a consistent, AI-powered understanding of customer risk. 

Wrapping Up

Now you know how to predict customer churn with the tools available inside Salesforce. More importantly, you know how to trigger re-engagement emails Salesforce marketers can trust to win customers back.

At Mavlers, we’ve seen how powerful this combination can be across industries and use cases.

Whether you’re just starting to explore Einstein or looking to optimize your existing churn workflows, our team is here to help you take the next step. Book a 30-min call with one of our Salesforce Marketing Cloud experts today!