From items to users: Rebuilding Plaid's API in flight
Plaid rebuilt its API around a new user-based model to better support cross-product financial workflows while maintaining its existing item-based infrastructure for backward compatibility. The company migrated hundreds of millions of users without disrupting integrations, unifying disparate user models from acquisitions and organic growth. This shift enables more holistic financial insights while preserving privacy and supporting use cases that require a unified view of user identity across accounts.
- ▪Plaid's original API was built around 'items,' representing a single financial account connection scoped to a specific customer.
- ▪Growing demand for identity verification, cash flow analysis, and cross-account insights necessitated a shift to a user-based API model.
- ▪Plaid consolidated multiple internal user models from products like Credit, Cognito, and Protect into a unified user primitive.
- ▪The migration was executed without breaking existing integrations, supporting over 7,000 customers on the legacy item-based system.
- ▪The new user-based API enables end-to-end workflows across products like Identity Verification, Auth, Credit, and Protect while maintaining privacy and data governance.
Opening excerpt (first ~120 words) tap to expand
From items to users: Rebuilding Plaid’s API in flightHow we rebuilt Plaid’s API around a new primitive — and migrated hundreds of millions of users to it without breaking a single integrationSam Subramanian16 min read·3 hours ago--ListenShareBy Sam SubramanianIn our first decade, Plaid’s products were centered around data aggregation — connecting bank accounts to financial applications. Plaid would extract raw data from banks, normalize it, enrich it, and send back a consistent view of balances and transactions. That powered things like account linking, personal budgeting, real-time balance checks, and payments.The core concept underneath all of this is the item.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at Medium.