Alternative Data is information outside a traditional credit bureau file that lenders use to assess whether someone will repay, including rent and utility payments, bank account cash flow, payroll data, telecom bills, e-commerce sales or, for businesses, payment-processor volumes. It matters most for applicants with thin or no credit files, who are otherwise declined or priced as high risk. Account cash flow, pulled with the customer's consent through an AISP connection, is usually the most predictive source. In the EU, GDPR, the second Consumer Credit Directive and the AI Act restrict which data can be used, so a buyer of a lender or scoring business checks the legal basis for every model input.
Who needs alternative data credit scoring
The US Consumer Financial Protection Bureau estimated in 2015 that 26 million Americans, about one in ten adults, were "credit invisible" with no file at the national bureaus, and a further 19 million had files too thin or too old to produce a score. Recent migrants, young adults and the self-employed fall into these groups across Europe too, which is why neobanks and credit scoring start-ups built models around bank transaction data.
Alternative data examples in lending
The most predictive source is usually account cash flow, obtained with the customer's consent through an AISP connection. It shows actual income, rent, existing loan repayments, gambling spend and how often the balance goes negative. For small businesses, merchant acquiring and marketplace sales data drive revenue-based financing and merchant cash advances, where repayments are taken as a percentage of daily card takings.
An illustrative case: an applicant with no bureau file earns €2,800 a month, pays €900 rent on time for 24 months and has never gone overdrawn. A bureau-only model declines; a cash-flow model can approve €3,000 over 12 months with a repayment of around €270, under 10% of net income.
Other sources are weaker and riskier: device data, email age, social media and psychometric tests. Their predictive value varies, and they create fair-lending and data protection exposure.
Alternative data regulation in the EU
Three regimes now limit what can be used. GDPR forbids processing special categories, such as health or ethnic origin, without a specific legal basis, and gives applicants the right to contest decisions made solely by automated means. The second Consumer Credit Directive (EU) 2023/2225, which member states must apply from 20 November 2026, limits creditworthiness assessments to relevant and accurate information on income and expenses, excludes special-category data and states that social networks are not an acceptable external source.
The AI Act classifies AI systems that assess creditworthiness of natural persons as high-risk. After the Digital Omnibus on AI entered into force on 27 July 2026, those obligations apply from 2 December 2027, covering risk management, data governance, logging and human oversight.
What it means in a fintech deal
For a buyer of a lender or scoring business, the model's data inputs are a liability check. Diligence should confirm the legal basis for each data source, whether any feature relies on social media or proxies for protected characteristics, and whether historical default rates were measured on loans the model actually approved.
See also: Big Data.