subscription-detection

Recurring charges are identified by matching same merchant, regular frequency, amount tolerance within ±5%, minimum 3 occurrences, and recency within the expected interval plus 7 days. The Skill exports merchant, frequency, and amount data to subscriptions.csv.

Finding subscriptions involves scanning transaction history for recurring patterns across merchants. The Skill detects frequency variations and amount changes, then exports merchant, frequency, amount, first_seen, last_seen, and status to subscriptions.csv for management and planning.

Automating subscription detection involves applying detection criteria—same merchant, regular frequency, amount tolerance ±5%, minimum 3 occurrences, and recency within expected interval + 7 days—then recording results in subscriptions.csv for budgeting and planning.

Subscription detection applies across categorized bank transactions, using five criteria to identify recurring charges. It supports learning corrections via learnings/subscriptions.md to improve accuracy over time for budgeting and expense optimization.

Recurring charges may fail detection if amount variations exceed the ±5% tolerance, or if fewer than 3 occurrences exist, or if the transaction falls outside the expected interval plus 7 days. All five criteria must be met for successful detection.

False positives can be corrected using the learning-driven feature, which updates detection rules via learnings/subscriptions.md. This improves accuracy over time by incorporating user corrections into future subscription detection analysis.