Short answer
Spending anomaly detection compares each new transaction with your own normal pattern and flags what does not fit: a charge far above your usual amount for that category, the same charge twice, or a subscription whose price went up. Finman alerts you on unusual spending and detects recurring charges, so you review only the outliers.
What counts as an anomaly
An anomaly is a transaction that breaks your pattern. A $2,000 rent payment is normal if you pay $2,000 every month; a $34 charge from a merchant that usually takes $12 is the one worth a look. Size alone tells you little. Detection that flags every large number gets ignored within a week, so useful detection works out what is normal for each category and merchant and speaks up only when something departs from it.
- RecordedNew transactionImported from a bank file, a connected account or typed in
- CalculatedYour normalTypical amounts for this category, merchant and recurring charge
- CalculatedFlagThe charge sits far outside that normal, repeats, or costs more than before
- You decideReviewLegitimate, a mistake to dispute, or a subscription to cancel
Three kinds of anomaly worth catching
- 01An outlier in its category. A grocery run three times your usual size, a fuel charge far above your normal fill-up. Detectors typically compare the amount with the spread of your past amounts in the same category and flag the ones far outside it.
- 02A duplicate charge. The same merchant and the same amount within a few days. Often legitimate (two coffees), sometimes a double charge you can get refunded.
- 03A recurring price that crept up. A subscription that was $9 and now posts at $13. The amount is small, which is exactly why it goes unnoticed for months. Recurring-charge tracking is what makes this one visible.
What Finman does with your transactions
Finman lists cash-flow forecasting and alerts on unusual spending among its features. Around that, three other automations feed the same picture:
- Categories. Imported transactions get a category from the merchant category code or from keywords in the description, in English, Ukrainian and Russian. Your own rules sort what the guess leaves in Other with one Apply rules click. Clean categories are what make “unusual for this category” meaningful; how transaction categorization works explains the mechanics.
- Recurring charges. Finman detects recurring charges in your transactions and surfaces subscriptions you may have forgotten, so you can cancel what you do not use. See the subscription tracker or the wider subscription tracker comparison.
- Budget pace. Finman flags a budget once 80% of it is spent, or when the current pace will break it before the period ends.
On top of that, the AI CFO chat answers questions such as “Where did I overspend this month?” from your own accounts and transactions, and writes a monthly summary with spending insights. It runs on an OpenAI GPT model reached from Finman’s server, and nothing is sent to it until you open the AI page or send a message.
Find the charge you would have missed
Import a few months of bank history and see what stands out. Free during the open beta, no card.
Start freeWhat anomaly detection cannot do
- It is a spending check, separate from fraud protection. A flag means “unusual for you”. A flagged charge may be fine, and an unflagged one is not certified safe. Card fraud stays your bank’s job; call them if a charge is not yours.
- It needs history. In the first weeks there is little to compare with, so expect fewer and rougher flags until a few months of transactions are in.
- It sees only what is recorded. If a card is missing from your imports, its charges are invisible. Import every account you spend from.
A 10-minute monthly routine
- 01Import every account. Download the statement files (CSV, OFX or QFX) or let a connected monobank account sync, so nothing is missing.
- 02Fix wrong categories first. A grocery charge filed under “Shopping” skews both categories. Add a rule so it lands right next time.
- 03Go through the flags. Recurring price rises first, then duplicates, then category outliers.
- 04Ask one question. For example: “Which categories grew the most this month?” Then decide on one change, such as one subscription to cancel or one category to cap.
If the anomalies keep pointing at the same category, the issue is a habit more than a one-off charge; overcoming impulse spending covers what to do about that.
Frequently asked questions
How does spending anomaly detection work?
It builds a picture of what is normal for you, per category, merchant and recurring charge, from your transaction history, then flags new transactions that depart from it: an amount far above usual for its category, a possible duplicate, or a recurring charge that got more expensive.
Does Finman alert me about unusual spending?
Yes. Finman lists alerts on unusual spending among its features. It also detects recurring charges, flags a budget at 80% spent or when the pace will break it, and its AI CFO chat answers questions about your own transactions.
Is anomaly detection the same as fraud detection?
No. It flags what is unusual for your own spending pattern. A flagged charge can be legitimate and an unflagged one can still be fraud; contact your bank about any charge you do not recognise.
How much history does anomaly detection need?
A few months of transactions. In the first weeks there is little to compare with, so flags get more useful as your history fills in.
Do I need to connect my bank for anomaly detection?
No. It works on whatever transactions are recorded: files imported from online banking (CSV, OFX or QFX), a connected monobank account or manual entry. Import every account you spend from so nothing is missed.
Let the unusual charges find you
Categories, recurring-charge detection and unusual-spending alerts on your own data. Free during the open beta.
Get started free