Financial data intelligence company Spade is rolling out a transaction foundation model aimed at giving regional banks and mid-tier fintechs the type of predictive tools that major Wall Street players have spent “billions and billions of dollars” building internally.
The new Spade Predictions product builds on the platform’s core data-enrichment offering, which involves deciphering complicated and often incomplete transaction data. Now, the company is taking that intelligence further by looking into what those transactions say about a person, anticipating consumer behavior and uncovering fraud.
Co-founder and CEO Oban MacTavish compared the New York-based fintech's offering to an Amazon Web Services-like utility that will allow smaller banks and businesses to access what industry giants are building with their own enormous resources.
“You need to be able to compete, and to compete, you need to partner, and we can be that sort of best in class partner,” MacTavish told This Week in Fintech.
Oban MacTavish
MacTavish cited a who’s who of big name firms that have trained models on their own transaction data in recent months using millions of their own customers and transactions, using dedicated research teams. Predictions offer the same capability to firms of every size, he said.
“Plaid had their foundation model, Revolut talked about it, NuBank, J.P. Morgan, Capital One, Visa, Mastercard…the list goes on,” MacTavish said. “The most sophisticated financial institutions are deploying these models…and the middle market banks need to compete. And I think that's really where we're sort of meeting this market moment right now.”
Building Off The Original Product
Spade’s model has two distinct components. One relies on Spade’s Merchant Graph, matching incoming transactions to its record of 99.9% of U.S. and Canadian businesses in real time. The other learns what a person’s spending says about them.
The company, which raised $40 million in Series-B funding earlier this year, said its foundation model, when fine-tuned for a U.S. bank’s lending program, outperformed a traditional propensity model built for the same task by 35%. On the fraud-fighting front, Spade said its model looked back at eight months of card issuer transactions and, in at least one instance, would have flagged suspicious transactions five months before the issuer.
The data won’t include personally identifiable information, or PII, Spade says; and MacTavish told TWIF they don’t even need it to curate recommendations.
The company, which already counts Stripe, Citizens, Robinhood and PayPal as clients, has traditionally generated revenue on a per-API call basis. With this new offering, MacTavish said Spade is exploring “success-based pricing.” Fraud prevention or better product targeting would follow that model, while other customers might prefer a fixed or licensing fee.
“The beauty of this is that there's so many ways it can be applied, and that means you have to sort of sort of meet your customers where they are in that sense,” MacTavish said.
Infrastructure Providers Are Pivoting
Spade isn’t the only company pushing past its infrastructure roots by moving into credit risk and decisioning. Just last week, Plaid introduced LendScore 2 and LendScore Arc, an updated credit-scoring system and a tool powered by AI. It joined the larger fintechs and banks MacTavish previously cited that have deployed their own proprietary models.
Spade’s product is for the rest of the financial community that can’t afford to sink millions into building their own foundation models.
“So much of this is about democratizing access to this next wave of innovation,” MacTavish said.
(Disclosure: Plaid is the parent company of This Week in Fintech; it has no say in our editorial process.)
