Case Studies 

Acquisitions foundational test for regional bank

Background: A regional bank wanted to double their customer credit loan portfolio using a substantially increased marketing budget. We worked across functions in the bank to design and implement testing that would provide the foundational information required to optimize their approval and line/price decisioning.

What we did:

  • Designed multi-dimensional marketing and line/price assignment testing that randomly assigned varied treatments to like customers.

  • Developed test size to secure the statistically significant reads of KPI’s, while prudently managing size and costs.

  • The data and subsequent NPV model developed using this test empowers the client to pull the right marketing/product levers to optimize their strategies and invest their marketing dollars with confidence.

Price and underwriting optimization for installment lending firm

Background: An installment lending firm hired AQN to review their risk model and suggest improvements including bringing in alternative data sources. We increased their profit per customer by >85% by using insight-driven segmentation to make pricing and underwriting tweaks to their strategy.

What we did:

  • De-averaged their loan portfolio from 5 segments into 30 segments based on significant differences in segment performance.

  • Built an NPV model to ground profitability for each segment.

  • Identified pricing and underwriting changes to maximize NPV with less than 150 bps impact to approval rates.

  • Identified strategy changes required minimal incremental investments and were implementable within two months.

Strategy for small business lender

Background: A small business lender wanted to re-evaluate its core strengths and develop a go-forward strategy with emphasis on growth and through-the-cycle resilience.

What we did:

  • Performed SWOT analysis, assessing the client's capabilities relative to the market and competitors.

  • Identified tactical optimization opportunities across different time horizons, including technology and talent investments.

  • Partnered with management and investors to prioritize next steps based on cost, upside, and potential risks.

Secured card CLI for regional bank

Background: A regional bank wanted better performance out of a secured card. We performed an assessment of available data, then built a new segmentation that drove 4x conversions and 8x profitability while reducing total exposure risk levels.

What we did:

  • Reviewed bureau data, demographic data, and the regional bank’s proprietary customer data to generate a full-spectrum assessment of risk and performance for an older sample.

  • Used machine learning techniques to split the population into segments by performance and risk band, then estimated T-C profitability and loss rate scenarios for a variety of lines in each segment.

  • The client agreed that the final proposal was more profitable and lower risk than their existing program and requested immediate execution of the program. A small control holdout and several small line tests were included in the roll-out to give the client insights for future program optimization.

Unit economics for international card business

Background: An international bank wanted to implement an analytics-driven decision-making framework to acquire credit card customers. We provided the client with the technical infrastructure and business expertise required to make unit-economics based acquisition decisions.

What we did:

  • Built a data-set of historic data containing all relevant drivers of profitability, and built the infrastructure to refresh the data-set.

  • Updated client’s cost allocation methodology to be driven by customer behaviors.

  • Collaborated with management to choose meaningful segments based on differentiation in behavior and ability to take action.

  • Predicted key customer behaviors for new acquisitions in each segment so the client has profitability expectations for the future.

Latent growth mixture modeling segmentation for regional bank

Background: A regional bank hired AQN to optimize credit card product terms, marketing strategy and credit risk so that it could efficiently grow its portfolio. We used advanced statistical techniques to isolate segments of accounts with materially different performance over time. These segments were used to improve valuation predictions, suppress risky segments from marketing, and focus the client on the most profitable accounts.

What we did:

  • Used latent growth mixture modeling (LGMM) to evaluate key performance drivers like balances and purchase volume.

  • Used decision trees to identify which booking attributes best explained the differing performance isolated by the LGMM analysis.

  • Improved valuations by providing granular segments of accounts and used these valuations to make improved, granular decisions related to marketing, product terms and risk.