What is alternative credit data and how do lenders use it?

#Why do some people have no credit score?

A credit score is calculated from your credit report. If you have never had a credit card or a loan, there is nothing to calculate. In 2015 the US Consumer Financial Protection Bureau (CFPB) estimated that 26 million American adults were credit invisible, meaning they had no credit record at the nationwide credit bureaus, and another 19 million were unscorable because their records were too thin or too old⁠^[1]. Those estimates were later revised. A Federal Reserve publication from October 2025, citing a CFPB technical correction of June 2025, puts the number of 'unscoreable' adults at roughly 32 million: 7 million credit invisible (2.7% of adults) and 25 million with a thin file (9.8%)⁠^[2].

This is a catch-22. You need credit to build a credit history, and you need a credit history to get credit. Young adults, new arrivals to a country and people who have always paid in cash are often stuck here, even if they pay every bill on time. Alternative credit data is one attempt to break the loop.

#What counts as alternative credit data?

The Federal Reserve describes alternative data as information from non-traditional sources that could bolster or sometimes substitute for traditional credit data, with benefits and risks for consumers⁠^[2]. When the CFPB asked the public about it in 2017, the examples it listed included mobile phone bills, rent payments, utilities and electronic transactions such as deposits, withdrawals and transfers⁠^[4].

#How do lenders get and use alternative data?

There are two main routes. In the first, the data is reported to a credit bureau and becomes part of your file, so ordinary scores can see it. In the second, you give a lender permission to read your bank account data directly, often through open banking connections, and the lender analyses your cash flow: money coming in, money going out and what is left. In December 2019 five US agencies, including the Federal Reserve and the CFPB, said that alternative data such as cash-flow data may expand access to credit and help some borrowers obtain more favourable pricing⁠^[5].

A real example comes from the US mortgage market. In August 2021 Fannie Mae announced that from 18 September 2021 its Desktop Underwriter system would, with the applicant's permission, automatically identify recurring rent payments in bank statement data⁠^[3]. Fannie Mae said fewer than 5% of renters had their rent payments reported on their credit bureau report, and that in a sample of applicants who had not owned a home in the past three years and did not receive a favourable recommendation, 17% could have received one if their rent history had been considered⁠^[3].

#Can machine learning on new data predict defaults better?

Sometimes, in some conditions. A Bank for International Settlements (BIS) working paper studied a Chinese fintech lender's data from 2017 and found that a model using machine learning and non-traditional data predicted losses and defaults better than traditional models when credit conditions suddenly worsened⁠^[6]. The same paper found that the advantage shrank for borrowers with a longer credit history⁠^[6]. In other words, new data helps most where old data is missing, which is exactly the thin-file problem.

#What are the risks of alternative credit data?

When the CFPB opened its 2017 inquiry, it flagged several risks: data that is inconsistent, incomplete or wrong; privacy concerns about how it is collected and used; the chance that some data acts as a stand-in for race, ethnicity or gender and leads to discrimination; and the difficulty consumers may have in understanding decisions or correcting errors in non-traditional data⁠^[4]. The 2019 interagency statement makes the same point from the lender's side: firms must check how fair lending law, the Fair Credit Reporting Act (FCRA) and rules against unfair, deceptive or abusive practices (UDAAP) apply before using new data⁠^[5].

One protection does not change when the data is new. Under US Regulation B, a lender that declines you must give specific principal reasons, not just say you failed to reach a qualifying score⁠^[7]. If a decision used alternative data, those reasons should still point to the factors that mattered.

#How does alternative data connect to BNPL and embedded lending?

Many newer lenders, including buy now, pay later providers and the platforms behind embedded finance, make quick decisions with their own data rather than a full credit report. The BIS has noted that big tech lenders such as Mercado Libre and Ant Financial say they use more than 1,000 data series per loan applicant⁠^[8]. That can widen access, but it also means your borrowing may not show up where other lenders look, which is a growing concern for credit reporting systems.