How Volix helped a company in the financial sector reduce churn and save R$1.8MM a year

Volix implemented an approach based on statistical modeling and machine learning to predict and reduce customer evasion with high precision. The project was structured in five main stages:

1. Deepening the Challenge

The first phase involved collecting and analyzing historical data to understand churn patterns and the factors that most impacted customer retention. Variables such as:

• Volume and frequency of transactions.
• Type of payment (credit vs. debit).
• Pricing variations and customer elasticity.
• Average turnover and purchasing behavior over time.

2. Data Preparation and Engineering

After identifying the critical factors, the databases were consolidated and processed to ensure quality and standardization. This process included:

• Cleaning up inconsistent data.
• Treatment of null values and outliers.
• Creation of new variables based on historical behavioral trends.

3. Development of the Predictive Model

The statistical approach chosen was Logistic Regression, which made it possible to estimate the probability of churn based on the explanatory variables. Complementary techniques were also tested:

• Random Forest

  to capture non-linear patterns in the data.

• GradientBoosting

  to improve predictive accuracy.

For each model, metrics such as Accuracy, Precision, Recall and ROC Curve were analyzed, ensuring the best possible performance.

4. Pilot Testing and Validation

With the predictive model developed, a pilot implementation was carried out on a controlled sample of customers. During this period, the team monitored:

• The effectiveness of forecasts.
• The response of customers to the retention strategies applied.
• The financial and operational impact of the model’s recommendations.

5. Implementation at Scale and Retention Strategies

After successful testing, the model was implemented across the entire customer base. With the insights gained, targeted retention actions were created, including:

• Personalized offers

  for customers at risk of churn.

• Pricing adjustments

  for segments that are more sensitive to the “take rate”.

• Engagement programs

  to increase loyalty and the frequency of transactions.