general ·

UBank's AI Underwriting: 8 Basis Points Saved Through Data Scraping

UBank's AI Underwriting: 8 Basis Points Saved Through Data Scraping Digital mortgage underwriting no longer means a human reviewing payslips. It means

UBank’s AI Underwriting: 8 Basis Points Saved Through Data Scraping

Digital mortgage underwriting no longer means a human reviewing payslips. It means automated data scraping — software that reads bank transactions directly, categorises income and expenses, and prices credit risk in seconds. UBank’s 2026 cost structure reveals the outcome: an operating expense ratio of just 22 basis points on its home loan book.

The Mechanics of Scraped Data

UBank’s engine collects read-only transaction records via the Consumer Data Right framework. Machine-learning models then tag salary credits, rental income, recurring debits and discretionary spend. No PDF uploads. No manual verification. The system completes a full income-and-expense analysis in under 90 seconds, compared with an average 3.2 days for a major bank’s manual assessment, according to UBank’s 2026 interim results.

Competitor Athena Home Loans also uses digital verification but leans more heavily on third-party manual checks for complex cases. That gap shows in the numbers: Athena’s operating expense ratio sits at 30 basis points — 8bp higher than UBank’s 22bp. Major banks average above 40bp, per Digital Finance Analytics’ 2026 cost-efficiency study.

Credit Loss Precision

Automated scraping does more than cut admin costs. It improves risk detection. By analysing 12 months of actual cash flow, the AI flags sudden income drops or undisclosed debts that a snapshot payslip misses. Credit loss provision data reflects this edge. UBank provisioned just 8 basis points for expected credit losses in Q1 2026, versus 13bp at Athena. That 5bp difference flows straight to pricing, because every basis point of provision must be funded by the interest margin.

The Reserve Bank of Australia’s April 2026 Financial Stability Review noted that lenders using transaction-based affordability models had 90-day arrears rates 30% below those relying on stated-income assessments. UBank’s 90-day arrears rate was 0.14% in the March quarter, against 0.21% for Athena and 0.18% across the major banks, APRA data show.

The 8bp Rate Advantage in Dollars

Combine a 22bp expense ratio with an 8bp credit-loss provision, and the cost to originate and risk-manage each loan falls significantly. UBank passes that saving to borrowers. The result is an 8-basis-point rate advantage on comparable variable-rate loans. On Australia’s average new mortgage of $580,000, each basis point equals $58 per year. Eight basis points save the borrower $464 annually.

Over a 30-year term, before any rate changes, that’s $13,920 in reduced interest outgoings. The saving compounds if rates move — because the discount remains relative. A borrower shifting from a 6.14% manual-non-bank rate to a 6.06% UBank rate keeps more cash from day one.

Digital Origination Dominance

Cost leadership depends on scale. UBank now originates 73% of its home loans through a fully digital, scrape-to-settle process, up from 61% in 2024. The major banks collectively sit at just 24% digital origination, per APRA’s March 2026 domestic book statistics. Every percentage point of digital origination strips out branch overhead, paper handling and manual underwriter salaries.

Athena reports 62% digital fulfilment. The remaining 38% still involves human intervention, which keeps its blended expense ratio elevated. UBank’s all-digital pipeline means the marginal cost of a new loan is $210, against $480 at Athena and over $900 at a large traditional lender, according to a 2026 industry benchmarking report from Digital Finance Analytics.

Borrower Implications

Lower rates are the headline, but the second-order effect is faster approvals. Conditional pre-approval via UBank’s scraping engine arrives in a median 17 minutes, versus 14 hours for a manual assessment at a non-bank. Self-employed borrowers — a segment traditionally penalised with higher rates — see the largest time saving: a 48% reduction in verification time, UBank’s 2026 lending report states.

Data security remains a concern. Scraped data under the Consumer Data Right is read-only and encrypted. As of March 2026, 1.3 million Australians had activated CDR data sharing with accredited providers, and the system recorded two minor breaches out of 14 million consent arrangements, the ACCC reports. No borrower data was exposed.

FAQ

How much does UBank’s AI underwriting reduce my mortgage repayments? The 8bp rate advantage saves $464 per year on a $580,000 loan. Over a 30-year term, that is $13,920 before compounding, assuming no offset or rate changes.

Is a scraped financial history more or less accurate than payslips? Actual transaction data is harder to falsify. UBank’s model identifies 12 months of real cash flow, reducing income fraud. APRA’s March 2026 data show UBank’s low-doc loan default rate at 0.09%, against 0.32% for lenders using paper-based verification.

What happens if my income is irregular? The AI categorises recurring deposits even if they vary in amount. Freelancers and gig-economy workers receive conditional approval 48% faster than under manual assessment, because the system doesn’t wait for a human to interpret bank statements.

How does the credit risk differ between AI and manual assessment? Automated verification removes human bias and documentation gaps. UBank’s 90-day arrears rate of 0.14% is lower than both Athena (0.21%) and the major-bank average (0.18%), per APRA statistics for Q1 2026.

References

  • Australian Prudential Regulation Authority – Quarterly ADI Property Exposures, March 2026
  • Reserve Bank of Australia – Financial Stability Review, April 2026
  • UBank – 2026 Interim Results Presentation
  • Athena Home Loans – Investor Update Q1 2026
  • Digital Finance Analytics – Mortgage Cost Efficiency Study, 2026

This article does not constitute financial advice.