Prism Data Raises $20M for Cash Flow Underwriting Models

A bank can hold a customer's deposits for years before those records become useful evidence in a lending decision. Prism Data sells the interpretation between the account and the credit team, helping institutions evaluate the income, spending and obligations already moving through a customer's financial life.

The New York cash flow underwriting company announced a $20M Series A on October 19, 2026, led by Story Ventures, with FINTOP, Core VC, Obvious Ventures, 1st & Main Growth Partners and Mana Ventures participating. The financing supports new credit models and broader deployment, as lenders look for ways to understand borrowers whose financial circumstances travel poorly through a conventional credit file.

What Prism Data's Series A will fund

Prism says the proceeds will support its AI infrastructure, data science and a new generation of CashScore models built on REFRACT, its newly introduced transaction categorization engine. The company plans to expand its work across consumer and small business lending in the US and Canada, with international ambitions over time. Alongside the $5M seed announced in 2023, Prism reports more than $25M raised; the new announcement leaves its valuation undisclosed.

The capital arrives at a business with an unusually layered history. Prism's official timeline traces model development to 2016 inside the team behind Petal, creation of Prism Data to 2020 and its emergence as an independent company to 2023. Treating all three milestones as a founding date would obscure the useful part: underwriting experience accumulated inside a lending business before becoming infrastructure sold to others.

Founder and CEO Jason Rosen previously co-founded and led Petal, according to his Consumer Bankers Association biography. Co-founder Jack Arenas is now a principal at Founder Collective, a separate present-day role. That origin gives Prism's financing a particular operating context, with tools developed around actual credit decisions now being offered across different institutions and portfolios.

For a buyer, that history raises questions more useful than the age of the startup. Which borrowers were represented in development, which credit products resemble the intended deployment, and how will the lender evaluate a model using its own lending outcomes? An established development history gives a credit team material to examine while it makes those decisions.

Why the meaning of a deposit matters

CashScore expresses risk derived from deposit activity as a 3-digit score. Prism describes a workflow in which clients bring account data directly or through aggregators, share de-identified information and receive analytical outputs for uses including approval, pricing and portfolio management. The lender retains the decision about what credit to offer and on what terms.

This creates a commercial opening between access to data and usable understanding. A balance alone cannot explain whether money came from recurring work, a refund or another borrowing obligation. Interpreting the activity gives the credit team a way to consider the circumstances surrounding that balance, rather than asking a single figure to explain the account holder.

Prism describes REFRACT as an AI engine trained on more than 30B transactions, intended to distinguish financial purposes that similar-looking bank entries can conceal. The investment in classification matters because a scoring model inherits the quality of the information beneath it. Recognizing a transfer accurately is part of deciding which evidence should influence an assessment of repayment capacity.

A lender evaluating such a system has practical work to do around the score itself. The institution needs to understand the input, test the output against its own population and decide how exceptions reach a human reviewer. A more detailed financial picture can improve the conversation, while also giving the people responsible for the decision more information to explain.

Cash flow underwriting meets the checkout clock

A July account from Synchrony makes the customer's side of this market visible. Synchrony describes offering some applicants the option to share bank activity when more information is needed for approval, while keeping the process within a checkout experience measured in seconds. That is a demanding setting for any additional analytical step: the shopper is waiting while the institution is still forming its view.

Synchrony says about 40% of applicants offered the option chose to share cash flow data, and roughly 40% of that group were approved. It expected 150,000 to 250,000 new accounts through the approach in 2026. Those are the lender's reported figures and forecast, rather than an allocation of results solely to Prism Data; Synchrony's own PRISM decisioning platform is a distinct system.

The example shows why adoption involves more than making another model available. Consent has to fit into the customer's task, information has to arrive in time and the credit policy has to say what the new evidence can change. A technically useful signal still needs an institution prepared to act on it while the applicant is present.

Prism reports that monthly API volume increased more than 100x over two years. That is evidence of reported activity on its infrastructure, with revenue, contract economics and credit outcomes requiring their own examination. Investors are backing a business whose commercial progress depends on lenders continuing to incorporate the analysis into their operations.

The model continues after the approval

Prism's Pulse product extends cash flow analysis into ongoing monitoring, including credit line management, risk assessment and collections workflows. Its product description includes changes in income, obligations and balances across existing customer relationships. That expands the possible relationship between a lender and its analytics supplier well beyond the first application.

The resulting opportunity also carries a wider operating responsibility. A signal about changing financial circumstances needs a policy for what follows: further review, an offer, a credit adjustment or assistance with repayment. Continuous information only becomes useful to a customer when an institution has decided how to respond to it.

FairPlay reported a fairness assessment of CashScore in September, including proxy testing, underwriting outcomes and scorability. The release describes the findings of that study; it provides a specific piece of validation evidence rather than blanket approval for every future model and deployment. As Prism funds the next generation of its technology, lenders will have new outputs to evaluate within their own governance.

Rosen's company is now financing both the machinery that interprets transactions and the models that consume the interpretation. At the institution buying it, the work continues inside individual lending policies and applicant files, where a newly understood deposit still needs somebody to decide what credit it can support.