$250K AUM Recovered via AI-Driven Churn Prediction System

$250K AUM Recovered via AI-Driven Churn Prediction System

Executive Summary

Cornerstone Wealth Management faced the common challenge of client attrition, impacting their AUM and revenue growth. To address this, they implemented an AI-driven churn prediction model built on Golden Door Asset's platform. This system proactively identified clients at high risk of leaving, enabling targeted intervention that successfully recovered $250,000 in Assets Under Management (AUM) that would have otherwise been lost.

The Challenge

Cornerstone Wealth Management, a growing RIA with $75 million in AUM, struggled with client attrition, a persistent issue impacting profitability and growth. While they had a strong client service model, their ability to proactively identify and address at-risk clients was limited. Existing methods relied heavily on manual reviews of client interactions and anecdotal evidence, proving insufficient in detecting subtle warning signs of potential churn.

Specifically, Cornerstone experienced an average annual attrition rate of 5% over the past three years. This meant losing approximately $3.75 million in AUM each year. While some attrition was unavoidable (e.g., due to client death or relocation), a significant portion was attributed to dissatisfaction or perceived lack of value. Identifying these clients proactively was crucial.

Their reliance on reactive measures, such as exit interviews and post-departure analyses, meant that valuable opportunities to retain clients were missed. For instance, a client with a $100,000 portfolio decided to move their assets to a competitor offering lower management fees. In retrospect, Cornerstone realized the client had expressed concerns about market volatility in previous conversations, a red flag that went unnoticed amidst daily operational tasks. Similarly, a client with $75,000 in retirement assets left due to perceived lack of personalized attention after a change in their advisor. These situations highlighted the need for a more sophisticated, data-driven approach to churn prediction. The lack of proactive insights resulted in missed opportunities to address client concerns, offer tailored solutions, and ultimately retain valuable assets. The manual process simply couldn't scale with Cornerstone's growth.

The challenge extended beyond just losing AUM; it also impacted Cornerstone's reputation and referral network. Dissatisfied clients were less likely to recommend Cornerstone to their peers, hindering organic growth. The financial impact of losing even a small number of high-net-worth clients was substantial, prompting Cornerstone to seek a more effective solution for predicting and preventing client churn.

The Approach

Cornerstone partnered with Golden Door Asset to implement an AI-driven churn prediction system. The strategic decision was to shift from a reactive to a proactive retention model, leveraging data analytics and machine learning to identify clients at high risk of attrition.

The approach involved several key steps:

1. Data Collection & Integration: The first step involved gathering and integrating relevant data from various sources. This included client demographics (age, income, net worth), transaction history (deposits, withdrawals, investment allocations), engagement metrics (website logins, email opens, meeting attendance), and communication logs (advisor notes, client feedback). This data was consolidated into a secure data warehouse. The key source was the existing Redtail CRM which held most if not all data.

2. Feature Engineering: Raw data was transformed into meaningful features for the AI model. For example, the frequency of client communication was calculated, recent investment performance was assessed, and changes in risk tolerance were identified. Special attention was paid to the rate of account activity: did this increase or decrease? Also looked at specific mentions within advisor notes in the CRM.

3. AI Model Development: Golden Door Asset's team of data scientists built a machine learning model using Python and the scikit-learn library. A gradient boosting algorithm was chosen for its ability to handle complex relationships and high dimensionality in the data. The model was trained on historical client data to identify patterns and predict churn probability. Multiple models were tested, including Logistic Regression, Random Forest, and XGBoost. Gradient Boosting provided the highest accuracy and recall.

4. Threshold Optimization: A critical step was determining the optimal threshold for triggering proactive intervention. A higher threshold would minimize false positives but risk missing at-risk clients, while a lower threshold would increase false positives but capture more potentially churned clients. Cornerstone conducted A/B testing with different threshold levels to optimize the balance between precision and recall. They settled on a threshold that identified the top 10% of clients most likely to churn.

5. Proactive Outreach Protocol: Once a client was flagged as high risk, a predefined outreach protocol was triggered. This involved a personalized email or phone call from the client's advisor, expressing concern and offering to address any issues. The advisor was equipped with relevant insights from the AI model, such as recent changes in portfolio performance or expressed concerns about fees.

6. Continuous Monitoring & Improvement: The AI model was continuously monitored and retrained with new data to maintain its accuracy and effectiveness. Cornerstone also regularly reviewed the outreach protocol to ensure it remained relevant and impactful.

Technical Implementation

The AI-driven churn prediction system was built using a combination of open-source technologies and proprietary Golden Door Asset algorithms. The core components included:

• Programming Language: Python (version 3.9) was used for data processing, model development, and API integration.
• Machine Learning Library: Scikit-learn was employed for building and training the gradient boosting model.
• Database: A PostgreSQL database was used to store and manage client data. Data was extracted using ETL pipeline built on Apache Airflow
• CRM Integration: An API was developed to integrate the AI model with Cornerstone's Redtail CRM system. This allowed for seamless data exchange and automated triggering of outreach protocols.
• Model Evaluation Metrics: The model's performance was evaluated using several key metrics, including:
  • Precision: The percentage of clients predicted to churn who actually churned.
  • Recall: The percentage of clients who churned that were correctly predicted by the model.
  • F1-Score: The harmonic mean of precision and recall, providing a balanced measure of model performance.
  • AUC-ROC: Area Under the Receiver Operating Characteristic curve, measuring the model's ability to discriminate between churn and non-churn clients.

The data integration process involved extracting client data from Redtail CRM via API, cleaning and transforming the data, and loading it into the PostgreSQL database. Feature engineering was performed using Python scripts, which calculated various client engagement and financial metrics. The gradient boosting model was trained using a combination of historical client data and engineered features. The model was then deployed to a production environment, where it continuously predicted churn probability for all clients.

The API integration with Redtail CRM allowed for automated triggering of outreach protocols. When a client's churn probability exceeded the predefined threshold, a notification was sent to the client's advisor via Redtail. The notification included relevant insights from the AI model, such as recent changes in portfolio performance or expressed concerns about fees. The advisor could then use this information to personalize their outreach and address the client's specific concerns.

The system also tracked the effectiveness of the outreach protocol. Data on client engagement after outreach, such as email opens, phone calls, and meeting attendance, was collected and analyzed to determine the impact of the intervention. This data was used to refine the outreach protocol and improve its effectiveness.

Results & ROI

The implementation of the AI-driven churn prediction system yielded significant results for Cornerstone Wealth Management:

• $250,000 in AUM Recovered: The system identified clients representing $250,000 in AUM at high risk of attrition. Proactive intervention prevented their departure, preserving these valuable assets.
• Reduced Attrition Rate: The overall attrition rate decreased from 5% to 3.5% within the first year of implementation. This translates to a reduction of $1.125 million in AUM lost annually.
• Improved Client Retention: The client retention rate increased by 1.5 percentage points, indicating improved client satisfaction and loyalty.
• Increased Advisor Efficiency: Advisors were able to focus their efforts on clients at highest risk of churn, improving their efficiency and effectiveness. Time spent on proactive retention efforts decreased by 20% due to the system's automated insights.
• Enhanced Client Relationships: Proactive outreach demonstrated Cornerstone's commitment to client service, strengthening relationships and fostering trust. Client satisfaction scores, measured through post-interaction surveys, increased by 10%.

Specifically, a client with a $125,000 portfolio was identified as high risk due to a decrease in website logins and an increase in withdrawal requests. Upon reaching out, the advisor discovered the client was considering moving their assets to a robo-advisor offering lower fees. By highlighting Cornerstone's personalized service and demonstrating the value of their financial planning expertise, the advisor was able to retain the client. Similarly, a client with $125,000 in retirement assets was flagged due to recent expressions of concern about market volatility. The advisor proactively addressed these concerns by rebalancing the client's portfolio and providing additional information about Cornerstone's risk management strategies, preventing their departure.

The ROI of the AI-driven churn prediction system was substantial. The $250,000 in AUM recovered generated approximately $2,500 in annual revenue (assuming a 1% management fee). Furthermore, the reduction in overall attrition rate resulted in a significant increase in revenue and profitability. The system paid for itself within the first quarter of implementation.

Key Takeaways

1. Proactive churn prediction is essential for sustainable growth: Relying on reactive measures is insufficient. A data-driven approach allows for early identification and intervention, preventing costly client attrition.
2. AI-powered tools can significantly improve client retention: Machine learning algorithms can analyze vast amounts of data to identify patterns and predict churn probability with high accuracy.
3. Personalized outreach is crucial: Once a client is identified as high risk, a personalized outreach protocol is essential to address their specific concerns and demonstrate the value of your services.
4. Continuous monitoring and improvement are key: AI models require ongoing monitoring and retraining to maintain their accuracy and effectiveness. The outreach protocol should also be regularly reviewed and refined based on client feedback and performance data.
5. Data integration is paramount: The success of any AI-driven solution depends on the quality and completeness of the data. Integrating data from various sources is crucial for building accurate and reliable models.

About Golden Door Asset

Golden Door Asset builds AI-powered intelligence tools for RIAs. Our platform helps advisors proactively manage client relationships, optimize portfolio performance, and enhance operational efficiency. Visit our tools to see how we can help your practice.