Beforepay releases new version of its AI-powered decision-analytics platform; prepares for overseas partnerships
Beforepay (ASX:B4P), a leading Australian fintech, unveiled its next-generation platform this week at the 2023 Singapore FinTech Festival. Beforepay showcased the Beforepay platform at the Singapore FinTech Festival as it begins exploring overseas partnerships with lenders and fintechs. The new version of the model features improved ensemble capabilities, seasonal factor adjustments, and enhanced transaction categorisation. Calculating 453 separate attributes for each first-time customer and 492 attributes for existing customers, it shows a meaningful upgrade in predictive power from the previous version.
The Beforepay platform has two main cores which can work in combination or independently. The first core powers insights for risk-based loan decisioning and limit-setting, while the second core enables a rapid and efficient end-to-end lending workflow. The platform’s decisioning core is built on the artificial-intelligence and machine-learning capabilities that have powered millions of loans in Beforepay's direct-to-consumer offering. The decisioning core is designed to combine a wide range of data sources in near-real-time to better understand customer risk and optimise customer lifetime value. A particular strength of the platform’s decisioning core is its ability to rapidly retrain on new data, enabling quick adaptation to new products and new markets once live. It will also allow for ensemble prediction, incorporating existing lending-model outputs to smooth the transition to more data-driven approaches.
The Beforepay end-to-end platform is built to onboard its customers, assess creditworthiness, predict customer lifetime value, set lending limits, and forecast defaults rapidly. The composable architecture will also enable the individual modules to be used separately, allowing potential partners to implement the platform’s AI-powered credit and risk models to enhance or replace existing scorecards.
Next-generation, data-driven loan analytics
The platform’s decisioning core is designed to ingest large volumes of data and use the data to make rapid, accurate decisions. Key capabilities of the platform’s decisioning core include:
Ingestion and pre-processing of multiple data sources into a consistent format to power artificial-intelligence training.
A range of risk models to assess customer risk from multiple perspectives, including multiple default paths (subject to a client’s access to categorised default data).
Ensemble architecture to allow custom-weighted combinations of different models, including legacy risk models.
A sophisticated limit-management module, including automated A/B testing of different limits to assess default elasticity and to maximise customer lifetime value.
Dynamic customer creditworthiness assessment, with fresh real time data enabling weekly rescoring and limit adjustments.
Elasticity modelling providing the ability to align specific risk settings to their risk appetite and drive their internal or external reporting. The model can explicitly set the balance of growth, risk and returns to deliver the preferred commercial outcome.
API access to core risk management components for integration.
Enhanced risk modelling capabilities
The decisioning core is built from the ground up on a base of artificial intelligence and machine learning (ML), instead of other platforms that attempt to retrofit ML variables into a scorecard architecture. This enables the testing of tens of thousands of features, and the selection of hundreds of features for the final model, each with individual statistically-derived inflection points and weightings. Multiple models can be combined through the ensemble architecture, allowing dedicated logic focused on specific paths to default.
The limit-management module allows for dynamic limit setting and limit adjustments, with automated testing of limit elasticity. This will combine with customer lifetime-value modelling to identify the limit that maximises the LTV of each customer on an individual basis given their risk score at any point in time.
The model architecture has been developed with a labelled dataset consisting of more than 3 million individual loans and more than 1.3 billion individual data points, a significant increase from the previous version. The model is trained out of the box and can incorporate potential partner data with variable weightings to tailor and localise the model. By default, the model is set to retrain every two weeks with data from the most recent loans, making it highly responsive to changes in macroeconomic conditions.
Speed and efficiency of end-to-end lending workflow
Beforepay’s end-to-end lending workflow platform has been built to handle large volumes of data quickly without any human intervention. It is currently running at scale in Australia, where a team of c. 30 employees using the platform issue more than 35,000 loans every week.
The platform has the ability to enrol most new customers in between five and ten minutes through a purely digital process, from app download to funds received, depending on product configuration, data collection times, and local payment rails. Thanks to the decisioning core’s ability to continuously update customer risk scores and dynamic adjust limits, returning customers can generally receive funds in less than one minute and new customers in as little as five minutes from phone unlock on Beforepay’s end-to-end platform.
Commitment to ethical lending
Beforepay is a mission-driven organisation dedicated to providing safe and affordable lending products to everyday consumers. Its direct-to-consumer business focuses on inexpensive, short-term, small-dollar advances to customers who otherwise might resort to revolving debt or predatory lending. This heritage underpins the platform’s focus on decisioning and sizing loans in a manner that reflects underlying customer needs, and also allows for additional business logic to support local requirements around responsible financial services.
Empowering financial innovation
Beforepay’s platform demonstrates the company’s leadership in ethical, sustainable lending, underpinned by its capabilities in AI, data science, and credit risk models, will enable flexible offerings for financial institutions once live.
"We’re excited to start conversations with potential offshore partners for international expansion, building on the success we’ve had in Australia" said Beforepay CEO Jamie Twiss. "With an AI-native risk-decisioning capability and the ability to ingest and process a wide range of data, our platform is well-positioned to support both traditional lenders and attackers looking to offer new products underpinned by modern decision science.”
The capabilities available in Beforepay’s platform have already driven Beforepay's rapid growth in Australia with its wage-advance product. Beforepay unveiled the platform at the 2023 Singapore FinTech Festival.
For more information about Beforepay visit www.beforepay.com.
