7
minute read
Nov 5, 2026

How Beforepay grew lending 10x, cut its credit loss rate by nearly 80%, and turned its margins positive

Custom credit risk models built around its own loan book helped Beforepay scale to about 40,000 loans a week, with average loan size up nearly 3x.

In short: Beforepay is a publicly traded Australian consumer lender that needed to grow lending without letting credit losses rise beyond its risk appetite. Carrington Labs built custom credit risk models around Beforepay’s own loan book and keeps monitoring and retraining them on new repayment data. Comparing Beforepay’s early results with the period since those models were in place, its credit loss rate went from 5.2% to 1.1% of the amount lent, annual lending volume from A$93m to A$963m, and risk-adjusted margin from negative 2.2% to 3.1%. Results reflect Beforepay’s portfolio and the periods shown.

Snapshot

  • Client type: Consumer lending fintech, publicly traded on the Australian Securities Exchange (ASX: B4P)
  • Market: Australia (all figures in Australian dollars)
  • Products: Unsecured Personal Loan and Pay Advance
  • Challenge: Increase loan volume and loan sizes while keeping credit losses within the lender’s risk appetite.
  • Carrington Labs models: Credit Risk Model, Credit Offer Engine, Cashflow Score
  • Integration type: API into Beforepay’s underwriting and origination workflow

About Beforepay

Beforepay is a publicly traded Australian consumer lender (ASX: B4P), recognized for ethical lending. It helps working Australians cover unexpected costs in a fair and transparent way. A car repair, a utility bill or the gap before the next paycheck can put pressure on a household budget, and Beforepay gives eligible customers a clear way to borrow through a simple app.

It offers an unsecured Personal Loan, a larger loan with a longer term, and Pay Advance, a short-term, small-dollar loan repaid in line with the customer’s pay cycle. Beforepay charges no late fees, no annual fees and no early repayment fees, and customers repay one loan in full before taking another, so borrowing stays manageable.

What challenge was Beforepay facing?

The balance between growing lending and controlling credit losses

Beforepay’s loans are built to be fair to the customer, which keeps its margin thin by design. That margin is what remains after credit losses, funding costs and servicing costs on each loan. With a fixed fee and a short loan term, there is little room to absorb credit losses. A handful of poorly judged approvals or oversized loans can erase the return on many well judged ones.

Beforepay’s earlier credit risk models already sized each loan to the customer’s ability to repay, but there was room to do it with greater precision. Without that sharper targeting, lifting volume or loan size meant taking on more credit risk than Beforepay could confidently measure. Holding back was the cautious option, but it left creditworthy customers with less than they needed.

Beforepay needed to approve more of the right customers and lend more to each, with every loan sized to earn the strongest margin without stretching the customer or pushing losses beyond its risk appetite. On a deliberately thin-margin product, that meant finding the amount where each loan delivers the most value to both the customer and the lender. It had to do this with a lean team, so its credit risk models needed to keep pace as the lending book grew and customer behavior changed. A one-time build would fall behind, so ongoing monitoring and refinement was essential, using new lending and repayment data to identify where risk scoring and loan sizing could improve. Because most of Beforepay’s lending is short term, repayment outcomes show up within weeks, so that data arrives quickly.

Results at a glance

  • Custom, lender-specific scoring. Carrington Labs’ Credit Risk Model, Credit Offer Engine and Cashflow Score gave Beforepay scoring and loan offers built around its own loan book.
  • Credit loss rate: from 5.2% to 1.1% of the amount lent.
  • Lending volume: from A$93m to A$963m a year, more than 10x.
  • Average loan size: from A$163 to A$456, nearly 3x, with each offer sized to the customer’s ability to repay.
  • Risk-adjusted margin: from negative 2.2% to 3.1% of the amount lent.
  • Continuous improvement. Beforepay’s models learn from new data fast. Between April 2025 and July 2026 alone, Carrington Labs trained and released ten models for Beforepay in 15 months, each retrained on new repayment data and tested against Beforepay’s own outcomes before going live.

What did Carrington Labs build for Beforepay?

Beforepay uses Carrington Labs to bring explainable cash flow intelligence into its underwriting and offer setting. The models are custom to Beforepay’s loan book, not a generic score, and they are retrained and improved as that book evolves. For every applicant, Carrington Labs supports Beforepay’s lending process across four steps:

  • Reading the data. Consumer-permissioned bank transaction data shows income consistency and spending pressure.
  • Scoring the applicant. The Credit Risk Model estimates the likelihood of default.
  • Setting the offer. The Credit Offer Engine selects the loan amount for each approved customer, within Beforepay’s risk thresholds and commercial objectives.
  • Returning the insights. Scores and recommendations flow into Beforepay’s own decisioning workflow.

Carrington Labs also stays with the models after launch. Its team monitors performance against Beforepay’s own lending results, identifies new signals in the data, and tests every change before releasing an updated model. That ongoing monitoring and refinement keeps Beforepay’s risk scoring and loan offers aligned as customer behavior, repayment patterns and the loan book change.

Beforepay retains full control of credit policy and lending decisions, and the models support its credit team rather than replacing it. The result is a granular view of risk for each customer and the confidence to lend more where the data supports it.

What changed for Beforepay?

Beforepay’s results moved on every measure a lender watches. This comparison sets its early results against its results since Carrington Labs’ models were in place, with each measure defined the way Beforepay reports it.

  • Credit loss rate: from 5.2% to 1.1%. This is the share of the amount lent, plus fees, lost to defaults and expected defaults, net of recoveries. It is closest to a US lender’s net loss rate, but it also includes expected losses provided for during the year.
  • Lending volume: from A$93m to A$963m a year. This is the total dollar value of loans originated, more than 10x.
  • Loans originated: from ~570,000 to ~2.1 million a year. That is about 40,000 loans a week.
  • Average loan size: from A$163 to A$456. This is total dollars lent divided by the number of loans, nearly 3x.
  • Risk-adjusted margin: from negative 2.2% to 3.1% of the amount lent. This is fee income less credit losses, funding costs and direct service costs, as a share of the amount lent. Beforepay reports it as net transaction margin.

Carrington Labs trains Beforepay’s models on new data quickly, and Beforepay’s lending makes that possible. Pay Advance loans are repaid within weeks, so results come back fast and the models can be retrained on fresh data almost immediately. That is how Carrington Labs delivered ten model releases in 15 months, between April 2025 and July 2026 alone. The updates refined how the models assess repayment behavior, introduced richer customer segmentation and added new signals, and each was tested against Beforepay’s own outcomes before going live.

That growth has allowed Beforepay to branch out into larger, longer-term Personal Loan products and value-add features such as Financial Pulse, which uses Carrington Labs’ Cashflow Score to give customers plain-English insights into their spending and help them build better habits.

"We built Beforepay to give working Australians a fair way to bridge the gap until the next time they get paid, and that only works if every loan is decisioned and sized with care.

Carrington Labs’ models have let us approve more of the right people and lend each of them an amount that works for them and for us, while our team stays in control of every decision. Because the models keep getting sharper, we can keep growing without losing sight of why we started.”

Jamie Twiss
Chief Executive Officer, Beforepay Group

What does Beforepay’s experience mean for other lenders?

Beforepay’s results show what can happen when a lender’s credit risk models keep learning alongside its business. Its lending volume grew more than 10x while its credit loss rate fell, as the models were monitored, retrained on new repayment data and refined release after release rather than built once and left alone.

These results reflect Beforepay’s own portfolio and loan products. Outcomes for other lenders depend on their data, products and borrowers, which is why Carrington Labs builds each model around the lender’s own book.

Beforepay’s lending operations provided the blueprint for Carrington Labs: a commercially minded, customer-centered way to turn bank transaction data into smarter lending insights and recommendations, tuned to each lender’s products, data and commercial objectives. Beforepay now runs on Carrington Labs models, and the results above reflect its portfolio over the years those models were built and refined. Beforepay and Carrington Labs are sister companies within Beforepay Group.

‍Who is this for?

Lenders running high volume, short term lending who want to grow approvals and loan sizes without loosening loss control.

Want to know more?

Carrington Labs builds custom credit risk models as a service for lenders, explainable and built around each lender’s own book. Beforepay’s growth shows what happens when models keep learning alongside the business: more of the right customers approved, each loan sized to what the customer can afford, and loss control kept in view.

Together, we can turn data into confidence, and confidence into growth.

Book a demo to see how Carrington Labs can build a model around your book.