Founded by ex-Googlers,

Upstart is the first lending platform to leverage artificial intelligence and machine learning to price credit and automate the borrowing process. Upstart has demonstrated unparalleled credit performance1 and the industry’s highest consumer ratings.2

In addition to its direct-to-consumer lending platform, Upstart provides technology to banks, credit unions and other partners via a “Software-as-a-Service” offering called Powered by Upstart.

in originations
expected in 2017
Lending is
centuries old,
Lending is centuries old,
but has changed little in recent decades. Almost all lenders use FICO-based models to decide who is approved for credit and at what interest rate. While simple and intuitive, these “scorecard” methods are limited in their ability to quantify risk.
More than
four in five
More than four in
five Americans

have never defaulted on a loan, yet less than half have access to prime credit.3

The implication is eye-opening: with a smarter credit model, lenders could approve almost twice as many borrowers, with fewer defaults.

AI/ML are upending
existing industries
and creating new ones. Whether self-driving cars, home assistants, or language translation, software that learns and improves on its own has moved from research labs to mainstream in just a few years.
Upstart is one of the first
to apply AI/ML to the multi-trillion dollar credit industry. Upstart goes beyond FICO, using non-conventional variables at scale to provide superior loan performance and improve consumers' access to credit.
The system
learning and optimizing in response to daily loan-level repayment and delinquency data.
Upstart has achieved separation.
This chart demonstrates how different versions of Upstart’s model would have predicted performance of more than 50,000 actual Upstart loans compared to how FICO would have scored them. Each dot represents about 400 loans. Over time, Upstart’s machine learning model is better able to identify risk, creating separation between the small number of high risk borrowers (top left) from the low risk borrowers (bottom right). Scroll to see how Upstart’s model has improved over time.
The Upstart model is learning quickly,
improving on its ability to identify creditworthy consumers. While the original model would have approved only 35 million Americans, the current model would approve more than 100 million, without a deterioration of credit.4
Num americans graph
Beyond credit scoring,
automation and modern data science can rewire the verification process, allowing Upstart to approve an increasing fraction of loans nearly instantly.
Loans auto graph
Virtually all lending
will be centered on AI/ML
Virtually all lending will
be centered on AI/ML
within a decade. While Upstart is focused today on the US consumer market, similar techniques can and will be applied to all credit markets.
Upstart has worked with
regulators since inception

to ensure it operates safely within the law.

AI/ML-based lending expands access to affordable credit by constantly finding new ways to identify qualified borrowers. Yet the model must avoid disparate impact, or statistical bias, that would be harmful to disadvantaged groups.

Upstart has worked transparently with regulators for more than 2 years and has consistently demonstrated that its platform doesn’t introduce bias to the credit decision. Furthermore, Upstart has developed operational and reporting procedures to ensure future versions of the model continue to be fair and unbiased.

Dave Girouard
Paul Gu
Anna M. Counselman
Alison Nicoll
Jeff Keltner
Jonathan Eng
Jungwon Byun
Barry Rafferty
Sanjay Datta
Jane Penner
Eric Schmidt
Mark Cuban
Marc Benioff
Join us as we bring a combination of modern data science, elegant design, and a commitment to learning and adapting to improve the lives of our borrowers.
1 6% annualized loss rates lowest among prime marketplace lenders based on Morgan Stanley securitization report.
2 According to a leading online loan aggregator.
3 Based on a Upstart/Transunion retro study & marketplace lending data.
4 Extrapolated from Upstart applicant data, adjusted for differences from general population using credit bureau data.