Spade Launches Spade Predictions Product for Banks and Fintech Companies

Spade has introduced Spade Predictions – a suite of products for banks and fintechs, providing access to transaction analytics that were previously developed primarily by the largest financial institutions. At its core is a fundamental model trained on payment behavior and matched with verified merchant data. Institutions can use it as is, further train it for specific tasks, or build their own models on Spade's software, infrastructure, and embeddings.

The launch comes as major banks and payment networks are increasingly creating AI models based on their transactional data. Such projects require extensive client histories, research teams, and significant investments, leaving smaller institutions with few resources for comparable solutions. Spade aims to make transactional AI more accessible to the entire industry.

Spade Predictions consists of two components: one identifies the business behind a transaction using verified merchant records, and the other interprets customer behavior, identifying needs and risks. This addresses a longstanding issue: card networks often limit the merchant name in transaction descriptions to 25 characters. The technology helps match customers with products, forecast spending, and detect fraud. Spade states that it does not receive personally identifiable information, and custom models use the institution's own identifiers.

Early tests are promising. A model further trained for a bank's credit program in the US outperformed a traditional propensity model by 35%. The version with Spade's merchant embeddings identified about 10% more future borrowers. In a fraud test, the model was evaluated on eight months of card issuer transactions: the system identified suspicious merchants early enough to block up to $3 million in fraudulent expenses. In one case, a signal appeared five months before the first block.

Predictions is the third suite on Merchant Graph, a database of verified businesses covering 99.9% of companies in the US and Canada. According to Spade, the graph matches merchants in real-time with over 99% accuracy in less than 40 ms. The database records what businesses sell, where they operate, their connections with brands, and how customers interact with them; locations are linked to parent brands and owners. The other two suites are Enrichment and Data Operations. Enrichment adds verified information about merchants, locations, and brands to transactions across all payment rails, including local, new, and third-party routed ones. Data Operations allows registering specific merchants and categories and obtaining one's own identifiers upon matches; Spade maintains matches during name changes, relocations, and new IDs.