Detect Recurring Transactions

That ordering isn't fussiness. A recurring charge is, by definition, one you've stopped noticing. If a single instance got dropped during extraction, the pattern can break just enough that you skim straight past it — and the 14.99 you meant to cancel keeps going out. Prove the statement is whole before you trust any pattern you spot in it.

Why bother? Three reasons. Subscriptions creep — accounts pick up small monthly charges nobody decided to keep paying, and they only stand out when you line them up. Recurring outgoings are the backbone of a budget; once you know what leaves every month before you've done anything, the discretionary spend is whatever's left. And regular income — the salary, the retainer, the rent received — is what makes a forecast worth the name.

What counts as a recurring transaction

Not everything that happens twice is recurring. The useful definition has two parts: a stable payee and a regular interval. A coffee shop you visit most weeks isn't a recurring transaction — it's frequent, but the amount and timing wander. A gym at 29.99 on the 3rd of every month is.

The common types worth naming:

• Subscriptions. Software, streaming, memberships, app charges. Usually card payments, often to the same merchant string each month — though the descriptor can shift slightly, which is what makes them fiddly to spot by eye.
• Standing orders. Fixed amounts you set up to leave on a schedule — rent, savings sweeps, regular transfers to another account. Same amount, same date, very predictable.
• Direct debits. Variable or fixed amounts a biller pulls — utilities, insurance, council tax, finance agreements. The amount can move month to month while the payee and rhythm stay put.
• Recurring income. Salary, pension, a retainer, rent received, regular client payments. The credit side of the same idea, and the half people forget to look for.
• Loan and finance repayments. Mortgage, car finance, a business loan. Large, fixed, and easy to take for granted precisely because they never change.

The reason eyeballing this is hard: a busy month buries the pattern. Forty card payments down a PDF, three of them the same subscription a fortnight apart, and the repetition just doesn't register. Sorting by payee is what makes it jump out — which is exactly the move the analyser automates once the statement is converted.

Prove the statement is complete first

Before you look for a single pattern, confirm nothing's missing. Every bank prints a running balance after each line — the true position of the account at that moment — so it works as a checksum. Take the opening balance, apply every credit and debit in order, and you should land precisely on the closing balance. Land there and the statement reconciles; you can trust what you build on it. Come up short and a transaction is missing or misread.

This bites hardest with recurring detection specifically. Lose one instance of a monthly charge to an extraction slip and you've turned a clean twelve-month pattern into eleven months with a hole — and an eleven-out-of-twelve pattern is exactly the kind a quick scan dismisses as "not really regular". The gap doesn't just cost you one row; it can cost you the whole finding.

Export Bank Statement runs the reconciliation check before you download. It extracts every line, walks the running balance from opening to closing, and confirms the totals tie out — or flags the statement rather than handing you tidy numbers with a quiet hole in them. Most converters extract and hope. For pattern work, hoping isn't enough, because a single dropped instance is the difference between catching a recurring charge and never seeing it.

How to detect recurring transactions, step by step

You can do this by hand — sort the rows by description, scan for repeats, note the interval and amount. It works on a quiet account. On a busy one it's slow and easy to get wrong, because the descriptors don't always match exactly and the eye tires. The faster route, on data you can trust:

1. Gather the statements. Download the PDFs from online banking — several consecutive months, ideally, since a pattern needs more than one instance to be a pattern. A scan or phone photo of a paper statement works too; it'll be read by OCR.
2. Convert them. Upload at /convert. The tool reads each line — date, description, money in, money out, balance — across every page.
3. Check the reconciliation result. The foundation. Confirm the totals reconcile before trusting anything downstream. A flagged statement means a gap got caught that would otherwise have broken a pattern.
4. Group by payee and read the recurring view. Open the analyser. It groups transactions by payee and surfaces the ones that repeat on a regular cadence — subscriptions, standing orders, direct debits, recurring income — with the interval and the typical amount. The expense analysis guide goes further on grouping the outgoings.
5. Act on what you find. Flag subscriptions you'd forgotten, confirm the income you expected actually landed every month, and carry the fixed outgoings into a cash flow forecast or a cash flow analysis.
6. Export. Take the clean .xlsx for your own working, or the CSV in the native bank-import format if it's headed for accounting software.

A single month runs in well under a minute, which is the point — it's quick enough to do every month instead of once a year, and recurring charges are exactly the thing you want to catch early.

A worked example

A small business, one quarter, around 300 transactions a month across a nine-page PDF. When I footed the first month by hand, page six didn't tie out — two rows had collapsed into one during a copy-paste, and one of them was a monthly software subscription. By eye, that subscription now showed up in only two of three months. Easy to write off as a one-off rather than the recurring charge it was.

Run the same PDFs through the converter and the reconciliation check flagged the month in seconds. Once the missing row was restored, grouping by payee told the real story: five recurring subscriptions running to about 340 a month between them, two of which nobody could account for; a standing order still leaving for a service that had been cancelled; and — on the credit side — a retainer that had quietly arrived three days later each month, worth knowing before it tightened cash in the second week. None of that is visible in a closing balance. All of it is obvious once the data is verified and grouped.

What this is and isn't

Worth being precise. Export Bank Statement converts your PDF statements, verifies they reconcile against the running balance, and surfaces recurring transactions through the built-in analyser after conversion. The honest caveat: detection runs on the statement you convert, not on a live connection to your bank. It doesn't read an ongoing bank feed or push anything into your ledger through a certified bank-feed API — you convert, the analyser groups what's there, and if it's headed for Xero, QuickBooks or Zoho Books you import the CSV, which lands as reconcilable statement lines. It's a tool you run yourself: the analyser shows the patterns; the decision to cancel, query or budget stays yours.