Neo Financial’s adjudication goals

• Replace the existing credit decisioning infrastructure with an enterprise decision engine which allows a high level of flexibility to credit risk managers in designing and testing multiple adjudication strategies and processes simultaneously
• Allow for faster time-to-market when deploying new decisioning logic or its changes
• Eliminate dependency on engineering and developer teams and give credit risk managers the ability to have end-to-end ownership of their processes

Why has Neo Financial chosen TaranDM

• “Platform-as-a-service” way of delivery – the decision engine runs at the client’s infrastructure (either cloud or on-prem), no sensitive data is shared with the vendor (customers’ personal details, etc.)
• Streamlined implementation & process migration, measured in weeks, and not in many months or years
• Modern technology stack – containerized, Python-based decision engine with microservices architecture
• Multiple out-of-the-box functionalities usually not available in other decision engines, such as Impact Report (simulation of decisioning logic’s changes against the status quo), batch decisioning, or Decision Optimizer

Implementation highlights

• TaranDM implemented and Neo Financial’s existing credit adjudication process migrated in less than 3 months
• Continuous knowledge transfer during the implementation & phase-in, along with TaranDM’s general customizability, has ensured that Neo Financial’s internal teams will not be overly dependent on us as the vendor going forward
• Higher flexibility when deploying new decisioning logic, for example:
• Credit product pricing now determined by the strategy and client historical profile/credit bureau record
• Specific strategies for selected partners which allow fine-tuning of approval rates/risk profiles and support of marketing actions and rollout plans
• Seamless A/B testing of strategies for various client segments

AREA42’s objectives

• Overarching goal: improve international trade by creating new and innovative technological solutions and a fluid, risk-savvy B2B trade ecosystem
• To achieve that, AREA42 has been looking for a decision engine with the following features:
• Ability to make quick, completely automated decisions
• Well-designed GUI for business users
• Modern technology stack and features such as easy data integrations, model deployment, out-of-the-box simulation layer, etc.

Why has AREA42 chosen TaranDM?

• Simple and flexible deployment of scoring models, including the support of multiple analytical environments for the model development
• Both real-time and batch decisioning-making and processing capabilities
• Easy maintenance of the decision engine
• Modern technology stack & “Python-native” nature of TaranDM
• Taran team’s credit risk and data science domain expertise which has ensured smooth & well-executed TaranDM implementation

Implementation highlights

• Development of a credit bureau adapter
• Deployment of the TaranDM Antifraud module:
• In-built data source which stores & automatically calculates concentration checks, cross checks and other aggregations which are crucial for fraud prevention
• Examples: number of loan applications from the same IP address over past 1h/24h/7d/30d; number of loan applications from the same device over past 1h/24h/7d/30d; etc.
• Includes hundreds of pre-defined aggregations and it’s customizable further (client can create additional aggregations)

Silk Bank’s objectives

Launching digital lending products with fully automated approval process

• Modernizing the bank’s credit risk decisioning process and infrastructure
• Ability to launch any new configuration of decision-making components in days

Why has Silk Bank chosen TaranDM

• Silk Bank chose TaranDM for its role in modernizing credit decisions and launching digital lending products. TaranDM's intelligent features ensure compliance and enhance financial precision. Beyond its impact on the approval process, TaranDM's ability to seamlessly integrate with various data sources positions it as a versatile solution for Silk Bank, ensuring a holistic and informed approach to decision-making. The platform will significantly improve our approval process by providing tailored product offers, backed by thorough credit bureau analysis. Its seamless integration with data sources, such as CreditBureau and the Revenue Service, underscores TaranDM as a comprehensive solution for Silk Bank. The platform's adaptability and robust features make it an instrumental asset in Silk Bank's journey towards modernized and efficient financial operations.
• In choosing TaranDM for our bank, we are leveraging a proven solution that has consistently delivered beyond expectations for reputable institutions. The client stories from TaranDM previous projects underscore the ability to not only meet but exceed strict timelines and facilitate the seamless migration of credit assessment processes. What distinguishes TaranDM is its adaptable deployment, showcased in prior project instances where it facilitated real-time data collection and operated seamlessly in a platform-as-a-service framework. This adaptability aligns with Silk Bank's need for a cutting-edge solution that integrates seamlessly with diverse data sources, empowering us to make informed decisions in real time. Additionally, TaranDM's commitment to knowledge transfer and customization ensures Silk Bank's teams are self-sufficient and not overly dependent on external vendors, which is a crucial aspect for long-term operational success. Additionally, TaranDM offers a user-friendly interface and a no-code graphical platform, empowering Silk Bank's team to adapt quickly to changing market demands and configure strategies independently. Its real-time processing capabilities ensure a swift and responsive customer experience, aligning seamlessly with Silk Bank's commitment to efficiency and customer satisfaction.
• In essence, TaranDM is not just a tool, it’s a strategic ally in our pursuit of excellence, combining cutting-edge data science with a user-friendly interface to drive innovation, efficiency, and superior customer satisfaction in Silk Bank's financial operations. This partnership positions Silk Bank ahead in the banking industry, empowering us to navigate the evolving landscape with agility and stay ahead in delivering unparalleled financial services.

Project highlights

• Implementation of the sophisticated income estimation logic to be compliant with the regulatory/internal requirements
• Enhancement of the overall approval process:
• TaranDM generates all product offers for which the customer is eligible
• That includes a refinancing offer, if relevant based on the credit bureau information
• Successful and continuous know-how transition to the Silk Bank’s team. Well before the production launch, they configured 2nd loan product by themselves in TaranDM’s intuitive, no-code GUI
• Easily integrating the approval process with several data sources, like Credit Bureau, Revenue Service, National Bureau of Enforcement, and National Bank of Georgia

Business challenge & objectives

• Product catalog in size of billions of items (hundreds of gigabytes of raw data), 8 mil.+ sellers, 1000+ merchants
• How to rank the products & maximize the conversion rates?
• How to run the product scoring process daily & efficiently, given the sheer size and complexity of the data?

Taran has exceeded all expectations

• A robust and scalable big data platform which requires minimal maintenance, and which has been used in production since 2019
• A state-of-the-art ML model for product scoring, which has led to 20% increase in revenue/conversions

The solution is leveraging the latest and best-performing big data technologies

• The pipelines include daily dump of product catalog, click logs and transaction logs into Amazon S3, processing the data using Spark on Amazon EMR, calculating aggregates and executing gradient boosting predictive models on PySpark
• Analytical playground connected with Git available on demand (in cloud, on click), servers provisioned daily using Airflow and Terraform
• Daily, Taran solution processes 600 GB+ data in less than an hour and calculates scores for billions of catalog items