Spending Patrol compares every new transaction to your historical baseline at that merchant. When a charge is 3× your typical amount it surfaces - useful for catching fraud, double-charges, and the gradual creep of "I didn't realize Uber cost that much now."
Uber posted $58 this morning. Your typical ride at that merchant is $12-16. Worth reviewing if this matches what you expected.
Most fraud detection waits for you to dispute a charge. By then the money is gone and you spend an afternoon on the phone. Most overspend detection lumps everything into a monthly category total - useful for budgets, useless for catching the one $400 Uber that should have been $40. The signal lives at the merchant level, not the category level.
For each merchant the agent maintains a running median + IQR of your historical amounts. New transactions get scored against the distribution: anything above 3× the median triggers a soft alert; anything above 5× triggers a hard alert with priority HIGH. Duplicate-charge detection runs in parallel - same merchant, same amount, within 48h, flagged as a likely double-post for review.
You take Uber to the gym three times a week, average $12-16. One Saturday a surge-priced ride home from a concert posts at $58. Without the agent it would blend into the monthly Transport line. With the agent you see it the next morning, decide it's legitimate (concert surge), and move on. If it had been a fraud charge you'd have caught it before the chargeback window closed.
Free on the Spark tier - no credit card required. Connect your bank read-only via Plaid; the agent starts watching from the first sync.
Try FinNudge free →No credit card · 5 minutes to connectAt least 3 transactions at the same merchant. Below that the baseline is unreliable - we'd be flagging "unusual" charges based on a sample of one, which is just noise. Most users hit the threshold within 30 days of connecting a bank.
Variable merchants (grocery stores, restaurants, gas stations) typically have wide IQRs, so the 3× threshold rarely fires. The agent is calibrated for fixed-ish merchants where a 3× delta really is unusual - coffee shops, gym memberships, ride-share, utilities.
No, complementary. Bank fraud detection looks at suspicious-merchant signatures and geographic anomalies. Spending Patrol looks at amount anomalies vs your personal history - so it catches things the bank thinks are legitimate (genuine Uber, genuine Amazon) but are unusual for you.