Design Financial Data Aggregation

Start by placing yourself correctly in the stack, because the design depends on it. You are the app a person opens to see where their money went, the layer above the plumbing that reaches into twelve thousand banks. Reaching every institution, holding the credentials, and scraping the flaky ones is a solved and specialized job, and you buy it from an aggregator like Plaid or MX rather than rebuild it. What the aggregator hands you is a firehose of raw transactions from a handful of different sources, in a handful of different shapes, arriving on their schedule and not yours. Your job begins there. Take that mess, fold it into one clean per-user ledger where every transaction appears exactly once and carries a category a person believes, and keep it current cheaply.

Two forces shape every later choice, and the first is that your inputs are heterogeneous and not yours. You may sit on top of more than one aggregator plus a few direct bank feeds, each with its own token model, its own field names, and its own idea of what a transaction looks like. The second, and the one that decides whether the product is any good, is categorization. A ledger of uncategorized transactions is a bank statement, which the user already has. The value you add is the sentence "you spent $340 on groceries this month," and that sentence is only worth anything if the category is right. So lead with two things: an abstraction that hides which source a transaction came from, and a categorizer people can trust.

The Intuit angle

Intuit's design round rarely names this prompt, so bring connect-and-categorize as product-domain color rather than expecting to be handed it. If the interviewer opens the fintech door, this is the design that shows you understand what QuickBooks, Credit Karma, and the old Mint do under the hood. The honest read from candidate reports is that Intuit's system design leans generic, and the real design signal is the Craft Demo. So treat this as depth you can reach for rather than a guarantee.