Lenders operating outside conventional banking use different information to assess borrowers, which produces different outcomes and different questions.
Alternative data
Cash flow, transaction history, rent and utility payments.
Which can demonstrate creditworthiness where a thin credit file does not.
Regulators have encouraged some of these uses as expanding access.
Cash flow underwriting
Assessing income and spending patterns from bank data.
Which requires access to accounts, generally through data-sharing arrangements.
It can identify affordability more directly than credit scores do.
Automated decisions
Models producing decisions without human review.
Which is fast and raises questions about explanation and recourse.
Adverse action notice requirements apply regardless of how a decision was made.
Fair lending obligations
Prohibitions on discrimination apply to all lenders and to all models.
Which includes disparate impact analysis in several regulatory frameworks.
Model testing for discriminatory outcomes is a supervisory expectation.
Proxy variables
Data correlating with protected characteristics.
Which can produce discriminatory outcomes without any protected characteristic being used.
This is a documented risk in machine learning applications generally.
Explanation requirements
Borrowers denied credit are entitled to reasons.
Which complex models make harder to produce meaningfully.
Regulators have stated that model complexity does not remove the obligation.
Bank partnerships
Non-bank lenders originating through partner banks.
Which raises questions about which jurisdiction's rules apply.
This structure has been the subject of litigation and regulatory attention.
For borrowers
Compare total cost including all fees, check the lender's licensing, and understand what data access is being granted.
This is general description rather than a recommendation about any lender.
Data sharing arrangements
Access to bank account data through defined interfaces.
Which regulation in several jurisdictions now governs explicitly.
Consumers have rights over what is shared and for how long.
Screen scraping versus interfaces
Older methods requiring credentials against modern permissioned access.
Which differ substantially in security.
Regulatory direction has been toward interfaces and away from credential sharing.
Small business lending
Alternative underwriting using payment processor and accounting data.
Which has expanded access for businesses without conventional records.
Cost of these products varies enormously and is not always expressed as an annual rate.
Disclosure of cost
Requirements to express financing cost in comparable terms.
Which several jurisdictions have introduced for commercial financing.
For borrowers
Convert every offer to an annualised cost before comparing, and check the lender's licensing.
Buy now pay later
Short-term instalment credit at the point of sale.
Which has been brought within credit regulation in several jurisdictions.
Reporting to credit bureaus has increased.
Earned wage access
Accessing wages before payday.
Which is marketed as not being credit and is treated as such by some regulators.
Fees expressed as an annualised rate can be very high.
Consumer protections
Disclosure, fair lending and collection rules apply according to product classification.
Which is why classification disputes matter.
Products marketed as not being credit attract particular regulatory attention.
Data rights
Rights to access, correct and delete personal data.
Which apply in several jurisdictions.
The borrower's checklist
Total cost annualised, licensing status, data access granted and what happens if a payment is missed.
Why alternative data expanded access
Conventional credit files exclude people who have not borrowed, including many who could repay comfortably.
Cash flow and payment history data address that directly, which regulators have generally encouraged while watching for discriminatory effects.
The borrower's position
Faster decisions, wider access, and products whose cost is frequently harder to compare than conventional credit.
Complaints and recourse
Regulators accept complaints about lenders and publish complaint data.
Which is searchable by company.
Complaint volumes and resolution patterns are informative before borrowing.
Licensing checks
State regulators maintain registers of licensed lenders.
Which takes minutes to search.
Unlicensed lending is a warning sign in itself.
The summary
Different data, faster decisions, wider access and products that require converting to comparable terms before assessing.
This is general description rather than a recommendation about any lender.
Model risk management
Testing, validation and monitoring of lending models.
Which supervisors expect regardless of the technology used.
Documentation of how a model works is a regulatory requirement.
Adverse action reasons
Specific reasons must be given when credit is denied.
Which applies however complex the model.
Regulators have stated explicitly that model opacity is not a defence.
The consumer's practical position
Request the reasons, check your credit file for errors, and compare total annualised cost across offers.
All three are free and all three change outcomes.
A closing note
Alternative underwriting genuinely expanded access to people conventional scoring excluded.
It also produced products whose cost is harder to compare, which is why converting everything to an annualised figure before deciding matters more than it used to.
Where to check
State licensing registers, regulator complaint databases and the loan agreement itself.
Fifteen minutes across all three before borrowing anything substantial.
A general caution
This describes how the system works and is not personalised financial advice.
Individual circumstances differ substantially, rules change, and anything consequential warrants a qualified professional who knows your situation.