Santos Financial: AI Predicts & Prevents 10% Client Churn

Santos Financial: AI Predicts & Prevents 10% Client Churn

Executive Summary

Santos Financial, a growing RIA firm, faced a persistent challenge of unpredictable client attrition, losing valuable assets under management (AUM) due to reactive retention efforts. To combat this, Dr. Elena Santos implemented an AI-powered model to identify at-risk clients based on activity patterns, communication history, and evolving market conditions. The result was a 10% reduction in client churn, translating to a $2 million increase in AUM retention and significantly improved client satisfaction.

The Challenge

Santos Financial, managing approximately $20 million in AUM, experienced an average client churn rate of 5% annually. While seemingly modest, the unpredictable nature of these departures created significant operational headaches. Losing a single high-net-worth client with a $500,000 portfolio could instantly wipe out weeks of new business development efforts. More concerningly, the firm noticed that a significant portion of these departures occurred with little to no warning, often attributed to "lack of engagement" or "finding a more suitable option."

Before implementing AI, Santos Financial relied primarily on reactive retention strategies. Upon receiving notice of a client's intent to withdraw funds, the team would initiate a series of phone calls, personalized emails, and even in-person meetings, offering tailored solutions and emphasizing the firm's value proposition. However, these last-minute interventions proved largely ineffective, with a success rate of less than 20%.

The existing system lacked the ability to proactively identify clients at risk of churning. While advisors tracked client interactions and attempted to gauge satisfaction levels, the sheer volume of client data and the subtle nuances of client behavior made it difficult to discern patterns and predict departures. This resulted in a reactive, firefighting approach to client retention, leaving Santos Financial vulnerable to unexpected losses and hindering their ability to achieve sustainable growth. Furthermore, the firm estimated that the reactive retention efforts consumed approximately 15% of advisor time, time that could be better spent on acquiring new clients and providing proactive service to existing ones. Losing just 3 clients with average AUM of $300,000 each represented a $900,000 loss, directly impacting profitability and growth targets. The unpredictable nature made financial forecasting inaccurate and resource allocation difficult.

The Approach

Dr. Santos recognized the need for a more proactive and data-driven approach to client retention. After researching various solutions, she opted to build an AI-powered model to predict client churn, focusing on identifying leading indicators of dissatisfaction or disengagement. The strategic decision framework involved several key steps:

1. Data Collection and Preparation: The first step involved gathering comprehensive data on existing clients. This included:

   • Demographic Data: Age, income, investment experience, risk tolerance.
   • Transaction History: Frequency of trades, types of investments, deposit/withdrawal patterns.
   • Communication History: Number of emails, phone calls, meeting attendance, response times.
   • Website Activity: Logins, page views, content downloads.
   • Market Data: Performance of client portfolios relative to benchmarks, relevant economic indicators.

   All data was anonymized and aggregated to ensure client privacy and comply with data protection regulations.

2. Model Development and Training: Using historical client data (covering the past 5 years), the team trained a machine learning model to identify patterns and predict which clients were most likely to churn. The chosen algorithm was a gradient boosting machine (GBM), known for its ability to handle complex datasets and identify non-linear relationships. The dataset was split into training (80%) and testing (20%) sets to evaluate model performance.

3. Feature Engineering: Key features were engineered to improve model accuracy. For example, a "portfolio underperformance ratio" was calculated by comparing the client's portfolio return to a relevant benchmark (e.g., S&P 500) over a specific period (e.g., 3 months). Another key feature was "communication decay," measuring the decline in communication frequency over time.

4. Threshold Optimization: A probability threshold was established to classify clients as "at risk." This threshold was carefully calibrated to balance the risk of false positives (identifying clients who were not actually at risk) and false negatives (failing to identify clients who were likely to churn).

5. Integration and Alerting: Once the model was trained and validated, it was integrated into Santos Financial's existing CRM system. When a client's probability of churning exceeded the predetermined threshold, an alert was triggered, notifying the client's advisor.

6. Targeted Interventions: Upon receiving an alert, advisors were equipped with specific information about the client's risk factors and recommended actions. These actions included:

   • Personalized Check-in Calls: Proactive calls to address any concerns and reaffirm the firm's commitment.
   • Portfolio Review: Offering a comprehensive review of the client's portfolio and suggesting adjustments based on market conditions and individual needs.
   • Educational Resources: Providing relevant articles, webinars, and other resources to enhance client understanding and engagement.
   • Special Offers: Offering exclusive access to new investment opportunities or discounted fees.

This strategic approach shifted the focus from reactive interventions to proactive engagement, allowing Santos Financial to address potential issues before they escalated into client departures.

Technical Implementation

The AI model was built using Python with the TensorFlow library, leveraging its capabilities for building and training machine learning models. The data preprocessing steps involved handling missing values, scaling numerical features, and encoding categorical variables.

The GBM model was trained using a cross-validation approach to optimize hyperparameters and prevent overfitting. The model's performance was evaluated using metrics such as precision, recall, and F1-score.

The integration with the CRM system was achieved using Zapier, a workflow automation tool. When the AI model identified a client at risk of churning, Zapier automatically triggered a series of actions, including:

• Creating a task in the CRM system for the client's advisor.
• Sending an email notification to the advisor with details about the client's risk factors.
• Updating the client's profile in the CRM system to reflect their "at-risk" status.

The calculations used within the model included:

• Sharpe Ratio: Used to assess portfolio risk-adjusted return. A declining Sharpe ratio could indicate increasing risk or decreasing returns, potentially prompting client dissatisfaction.
• Information Ratio: Used to measure portfolio performance relative to a benchmark, adjusted for risk. A negative or declining information ratio could signal underperformance.
• Engagement Score: A proprietary metric calculated based on client activity levels, communication frequency, and website interactions. This score provided a comprehensive measure of client engagement.

The model also incorporated economic indicators such as the Consumer Price Index (CPI), the unemployment rate, and the yield curve to account for the impact of macroeconomic factors on client sentiment and investment decisions.

Results & ROI

The implementation of the AI-powered churn prediction model yielded significant results for Santos Financial.

• Reduced Client Churn: The annual client churn rate decreased from 5% to 4%, representing a 20% reduction in attrition.
• AUM Retention: This reduction in churn translated to a $2 million increase in AUM retention. Prior to the AI implementation, losing 5% of $20 million in AUM resulted in a $1 million loss. Reducing that to 4% resulted in only a $800,000 loss. The difference of $200,000 per year over 10 years, conservatively estimated, is $2 million.
• Improved Advisor Efficiency: By proactively identifying at-risk clients, advisors were able to focus their efforts on targeted interventions, reducing the time spent on reactive retention efforts by 30%.
• Increased Client Satisfaction: The proactive engagement strategy led to improved client satisfaction scores, as clients felt more valued and supported. Client satisfaction scores increased by 15% based on post-intervention surveys.
• Increased Revenue: The retained AUM generated an additional $20,000 in annual revenue based on a 1% AUM management fee. This does not include the potential revenue from new assets invested by retained clients.

The ROI of the AI implementation was significant. The development and implementation costs were estimated at $50,000, while the annual benefit from reduced churn and increased revenue was estimated at $20,000 + a significantly improved staff utilization rate. The project paid for itself in less than 3 years, and delivered long term financial benefits while significantly reducing client and employee stress.

Key Takeaways

Here are some key takeaways for other RIAs considering implementing AI-powered client retention strategies:

1. Data is King: Accurate and comprehensive client data is essential for building effective predictive models. Invest in data collection and management processes.
2. Proactive is Better than Reactive: Shifting from a reactive to a proactive approach to client retention can significantly reduce churn and improve client satisfaction.
3. Personalization is Key: Tailor your interventions to the specific needs and concerns of each client. Generic outreach is unlikely to be effective.
4. Integration is Crucial: Integrate your AI model with your existing CRM system to streamline workflows and ensure that advisors have the information they need to take action.
5. Start Small and Iterate: Begin with a pilot project to test your model and refine your approach before scaling it across your entire client base.

About Golden Door Asset

Golden Door Asset builds AI-powered intelligence tools for RIAs. Our platform helps advisors identify at-risk clients and provide personalized, proactive service. Visit our tools to see how we can help your practice.