How Recurring Detection Works

Automatic Pattern Detection

NivoaFlow scans the last 18 months of your transaction history to find recurring patterns automatically. When enough transactions line up — three or more in most cases, or two for yearly bills — we flag them as a potential recurring pattern for your review.

Transfers between your own accounts are excluded from detection, so credit card payments and savings contributions do not clutter your bills list. See transfers and double entry.

How Matching Works

The detection engine groups transactions by normalised merchant name (so "Netflix" and "NETFLIX.COM" land together) and looks at three signals:

• Merchant — same cleaned-up merchant name
• Cadence — the median gap between transactions has to fall inside a known window:
  • Weekly: 6–8 days
  • Biweekly: 13–16 days
  • Monthly: 26–35 days
  • Quarterly: 80–100 days
  • Yearly: 330–400 days
  • Custom: anything else, only if the timing is otherwise very consistent
• Amount — values should be similar; small variations (tax changes, price tweaks) are tolerated

Confidence Scoring

Each candidate pattern earns a confidence score from 0 to 100, built from:

• How tight the timing is (lower variation, higher score)
• How tight the amounts are
• How many occurrences we have seen (more is better)

Patterns under a confidence of 40 are dropped. Two-occurrence yearly patterns are capped at 65 so they stay in the lower-confidence range until a third occurrence confirms them.

Reviewing Detected Patterns

Detected patterns appear on the Recurring page. From there you can:

• Confirm a pattern to add it to your tracked recurring transactions
• Edit the details if the detection got something slightly wrong
• Dismiss a pattern if it is not actually recurring

Staleness and Reactivation

After detection, NivoaFlow runs a staleness pass over every active pattern:

• One missed expected charge → pattern is marked possibly cancelled
• Two consecutive misses → pattern is marked cancelled and stops projecting
• A fresh matching transaction reactivates the pattern automatically

This keeps your forecasts and bill calendar accurate when you cancel a subscription or a price changes meaningfully.

Getting the Best Results

For the most accurate detection, keep your accounts connected and transactions flowing in regularly. The more history we have to work with, the better the pattern matching becomes.