Methodology
How the Spending Analyzer works
The Spending & Subscription Analyzer parses a CSV or PDF statement entirely in-memory on our server, categorises every transaction with a large-language-model classifier, and surfaces the result as an aggregated breakdown plus the top five sample transactions per category. Nothing is written to a database or to disk, no account is created, and the moment the response is sent your parsed data is garbage-collected.
Categories are mapped to seven standard buckets — Housing, Groceries, Dining & Coffee, Transport & Fuel, Utilities, Shopping, and Subscriptions — plus an "Other" residual. Recurring charges are detected by looking for the same merchant within a ±4-day window across two or more billing cycles, which is the industry-standard subscription-detection heuristic.
The surplus figure is simply total income − total spend for the period. Bill benchmarks are hard-coded by US region (Northeast, South, Midwest, West, Non-US) using 2025 Consumer Reports and CTIA medians; a bill is flagged when it exceeds 130 % of the regional typical for its category. The retirement-boost CTA passes the computed surplus into the Retirement Calculator via a query parameter so you can see the 30-year compounded impact of redirecting it into an index fund.