Predictive Attrition Modeling: 12% Reduction in Client Churn

Predictive Attrition Modeling: 12% Reduction in Client Churn

Executive Summary

Ferguson Estate Planning, a growing RIA struggling with client attrition, faced a reactive approach to retention, only identifying potential departures after clients initiated withdrawal requests. Golden Door Asset developed a predictive attrition model that analyzes client data to proactively identify at-risk clients, enabling personalized support and targeted interventions. This AI-powered solution reduced client churn by 12% within the first year, translating to an estimated $75,000 in saved revenue.

The Challenge

Ferguson Estate Planning, a firm managing over $50 million in assets for approximately 150 clients, recognized that client retention was crucial for sustainable growth. However, their client retention strategy was largely reactive. They typically became aware of a client’s intent to leave only after receiving a withdrawal request or transfer paperwork. This late notification severely limited their ability to address the client's concerns and proactively prevent churn.

In the previous fiscal year, Ferguson Estate Planning experienced a client attrition rate of approximately 8%. While seemingly manageable, the impact on revenue was significant. With an average client portfolio size of $333,333 and an advisory fee of 1% annually, each lost client represented a loss of $3,333 in annual revenue. The 8% attrition rate equated to approximately 12 clients departing, resulting in a total revenue loss of $40,000.

Furthermore, the costs associated with acquiring new clients to offset these losses were substantial. Industry averages suggest that acquiring a new client can cost anywhere from $2,000 to $5,000, depending on marketing efforts and referral programs. To replace the lost revenue, Ferguson Estate Planning would need to acquire roughly 12 new clients, potentially incurring acquisition costs of $24,000 to $60,000. This represented a significant drain on resources and hampered overall firm growth.

Beyond the direct financial impact, the reactive approach also affected staff morale. Investment advisors often felt frustrated when they were unable to address client concerns proactively, leading to feelings of powerlessness and decreased job satisfaction. The repetitive task of processing withdrawal requests and initiating the onboarding process for new clients also consumed valuable time that could have been spent on more strategic initiatives.

The firm's principal, Sarah Ferguson, recognized the urgent need for a proactive client retention strategy. She stated, "We were essentially playing defense, always one step behind. We needed a way to anticipate potential departures so we could engage clients and address their concerns before they decided to leave." The key challenge was identifying the specific factors that contributed to client attrition and developing a system for early detection.

The Approach

Golden Door Asset partnered with Ferguson Estate Planning to develop a predictive attrition model based on machine learning techniques. The initial phase involved a thorough data audit and needs assessment. We worked closely with the Ferguson Estate Planning team to understand their existing client data infrastructure, including their CRM (Client Relationship Management) system and financial planning software.

We identified several key data points that could potentially correlate with client attrition, including:

• Demographic Information: Age, location, income level, and marital status.
• Portfolio Characteristics: Asset allocation, investment performance, risk tolerance, and portfolio size.
• Account Activity: Frequency of account logins, website visits, transaction history, and communication logs.
• Financial Planning Metrics: Progress towards financial goals, retirement projections, and insurance coverage.
• Customer Service Interactions: Number of support tickets, satisfaction scores, and communication frequency.

Once the relevant data was identified, we cleaned, transformed, and preprocessed the data to ensure its quality and consistency. This involved handling missing values, correcting inconsistencies, and converting categorical variables into numerical representations suitable for machine learning algorithms.

Next, we explored various machine learning models to identify the one that best predicted client attrition for Ferguson Estate Planning's specific dataset. We considered logistic regression, decision trees, random forests, and support vector machines (SVMs). Through rigorous testing and validation, we found that a random forest model provided the most accurate and reliable predictions. The random forest model is an ensemble learning method that combines multiple decision trees to reduce overfitting and improve generalization performance.

The model was trained using historical client data from the past three years. We used 80% of the data for training and the remaining 20% for testing the model's accuracy. The model was trained to identify patterns and relationships between the input features and the target variable (client attrition).

Finally, we developed a user-friendly dashboard that presented the model's predictions to the Ferguson Estate Planning team. The dashboard provided a list of clients ranked by their probability of attrition, along with the key factors contributing to their risk score. This enabled advisors to prioritize their outreach efforts and focus on the clients who were most likely to leave.

The strategic thinking behind this approach centered around proactive engagement and personalized support. By identifying at-risk clients early, advisors could initiate targeted interventions, such as:

• Proactive Check-in Calls: Scheduling regular calls to discuss client concerns and address any unmet needs.
• Personalized Financial Planning Reviews: Reviewing clients' financial plans and making adjustments to reflect their changing circumstances.
• Educational Workshops: Providing educational resources and workshops on relevant financial topics.
• Value-Added Services: Offering additional services, such as estate planning assistance or tax optimization strategies.

The ultimate goal was to strengthen client relationships, build trust, and demonstrate the value of the firm's services.

Technical Implementation

The predictive attrition model was developed using Python with the scikit-learn library, a popular open-source machine learning framework. The following steps outline the technical implementation:

1. Data Extraction & Preparation: Data was extracted from Ferguson Estate Planning's CRM (Salesforce) and financial planning software (eMoney Advisor) using their respective APIs. The data was then loaded into a Pandas DataFrame for cleaning and preprocessing. Missing values were handled using imputation techniques (e.g., mean imputation for numerical features and mode imputation for categorical features). Categorical variables were encoded using one-hot encoding.

2. Feature Engineering: Several new features were engineered based on the existing data to improve the model's predictive power. Examples include:

   • Investment Performance Relative to Benchmark: Calculated the difference between the client's portfolio performance and a relevant benchmark index (e.g., S&P 500) to assess relative investment performance. This used a time-weighted return calculation.
   • Communication Frequency: Calculated the number of emails, phone calls, and meetings with the client over the past year.
   • Account Activity Score: Created a composite score based on the frequency of account logins, website visits, and transaction activity.
   • Fee Ratio: Calculated the ratio of advisory fees paid to assets under management to identify clients who might be sensitive to fees.

3. Model Selection & Training: A Random Forest Classifier was selected as the primary model due to its ability to handle non-linear relationships and its robustness to overfitting. The model was trained using 80% of the historical client data, with a stratified split to ensure that the training and testing sets had similar proportions of attrition cases. Hyperparameter tuning was performed using cross-validation to optimize the model's performance. The primary hyperparameters tuned were n_estimators (number of trees in the forest) and max_depth (maximum depth of the trees).

4. Model Evaluation: The model's performance was evaluated using various metrics, including:

   • Accuracy: The overall percentage of correctly classified clients.
   • Precision: The proportion of clients predicted to churn who actually churned (True Positives / (True Positives + False Positives)).
   • Recall: The proportion of clients who actually churned who were correctly identified by the model (True Positives / (True Positives + False Negatives)).
   • F1-Score: The harmonic mean of precision and recall, providing a balanced measure of the model's performance.
   • AUC-ROC: The area under the Receiver Operating Characteristic curve, which measures the model's ability to distinguish between churn and non-churn clients across different probability thresholds.

5. Dashboard Development: A user-friendly dashboard was developed using Tableau to visualize the model's predictions and provide insights to the Ferguson Estate Planning team. The dashboard included a list of clients ranked by their probability of attrition, along with the key factors contributing to their risk score. The dashboard also provided visualizations of the model's performance metrics and feature importance.

6. Integration: The model was integrated into Ferguson Estate Planning's existing workflow via a weekly batch process. A script automatically pulls the latest client data, scores it with the model, and updates the Tableau dashboard.

Results & ROI

The implementation of the predictive attrition model yielded significant positive results for Ferguson Estate Planning:

• Reduction in Client Churn: The client churn rate decreased from 8% in the previous year to 6.8% in the first year after implementing the model. This represents a 12% relative reduction in client churn.
• Saved Revenue: With an average client portfolio size of $333,333 and an advisory fee of 1%, each prevented churn translated to a savings of $3,333 in annual revenue. By reducing churn by 1.2%, the model helped Ferguson Estate Planning retain approximately 4 clients, resulting in an estimated $75,000 in saved revenue ($3,333 x 4 x ~5.5 years average client lifespan).
• Improved Client Relationships: The proactive outreach efforts enabled advisors to strengthen their relationships with clients and address their concerns before they escalated. Client satisfaction scores, measured through post-interaction surveys, increased by 15%.
• Increased Advisor Efficiency: By focusing on at-risk clients, advisors were able to prioritize their outreach efforts and make more efficient use of their time. Time spent on reactive churn management decreased by 20%, freeing up valuable time for business development and other strategic initiatives.
• Reduced Acquisition Costs: By preventing client churn, Ferguson Estate Planning reduced the need to acquire new clients to replace lost revenue. This resulted in an estimated savings of $8,000 to $20,000 in acquisition costs (based on an estimated $2,000 to $5,000 cost per new client acquisition).

The model also provided valuable insights into the key drivers of client attrition at Ferguson Estate Planning. The most significant factors identified by the model included:

• Investment Performance Below Benchmark: Clients whose investment performance consistently lagged behind relevant benchmarks were more likely to churn.
• Decreased Communication Frequency: Clients who had less frequent communication with their advisors were also more likely to churn.
• Lack of Engagement with Financial Planning: Clients who did not actively engage with their financial plans or update their goals were at higher risk of attrition.

These insights enabled Ferguson Estate Planning to refine their client service offerings and address the specific needs of their clients more effectively.

Key Takeaways

1. Proactive Retention is Crucial: Don't wait for clients to initiate withdrawal requests. Implement proactive strategies to identify and address potential attrition risks early.
2. Data is Your Best Friend: Leverage your existing client data to identify patterns and predict attrition. Invest in data infrastructure and analytics capabilities.
3. Personalization Matters: Tailor your outreach and support efforts to the specific needs and concerns of each client.
4. Focus on Key Drivers: Identify the factors that contribute to client attrition at your firm and develop strategies to mitigate these risks. Is it performance? Service? Fees? Address these directly.
5. Measure and Iterate: Continuously monitor the effectiveness of your retention efforts and make adjustments as needed. Regularly retrain your model with new data to maintain accuracy.

About Golden Door Asset

Golden Door Asset builds AI-powered intelligence tools for RIAs. Our platform helps advisors proactively manage client relationships and identify growth opportunities. Visit our client attrition prediction tool to see how we can help your practice.