Case Study: $18B Wealth Manager Reduces Customer Attrition by 15%

The client

The investment banking and capital markets arm of a major banking group, and one of the largest wealth managers in Turkey.

The challenge

The firm needed to understand customer behavior and make its retention efforts more predictable. Earlier AI projects run by data science teams had failed. The models were black boxes, so business teams could not act on them or improve them.

The approach

TAZI was installed on the firm's own servers to keep data in-house. The work ran in four phases:

1. Initial models. Priority variables included total assets transacted, number of transactions, commission rate and total commissions paid. The attrition definition was widened beyond account closures to include minimal-balance and inactive accounts.
2. Data quality. TAZI's Data Profiler assessed data relevance and quality. Historical data was reconstructed to reflect past conditions, with continuous quality checks.
3. Optimization with business users. Business KPIs such as commissions paid were built into model metrics. Feature engineering was automated and guided by business users, using more than 100 variables. Each prediction came with an explanation.
4. Production. Real-time dashboards were set up, and models were A/B tested before deployment, with minimal IT involvement.

Predictions fed retention actions across departments: calls, text messages, and alerts to sales representatives about at-risk customers and the effect on their performance.

Results

• 15% reduction in customer attrition
• $7.5M revenue saved per year from attrition-prevention actions
• 100+ variables used to model attrition
• Improved data quality through continuous control of data processing and storage
• Team skills from the project now applied to new AI initiatives

Over the last four years, the firm grew its customer base 7x and its company value 6x. The solution has been in production for five years.