How to Categorise Bank Transactions Automatically (Free)

Sorting a year of transactions by hand is the reason most budgeting attempts die in week two. Roughly a thousand rows, each needing a glance and a category, and the job has to be redone next month.

Our bank statement converter does the sorting as part of the conversion. Upload a PDF, get a spreadsheet where every row already carries a category.

How the matching works

It is a keyword rule list, not a machine-learning model. Each transaction description is checked against a set of merchant names and phrases, and the first matching rule wins.

That design has three consequences worth knowing:

• It is instant. There is no model to run, so a year of statements categorises as fast as it parses.
• It is predictable. The same description always produces the same category. You are never left wondering why last month's Tesco went to Groceries and this month's went to Shopping.
• It is auditable. If something lands in the wrong bucket, the reason is a keyword, not a black box.

The categories

Income, Groceries, Eating out, Transport & fuel, Shopping, Bills & utilities, Connectivity, Subscriptions, Insurance, Health & pharmacy, Banking & fees, Loans & debt, Savings & investing, Cash, Education, Travel, and Other.

Seventeen buckets is a deliberate middle. Five is too coarse to act on — you learn that you spend money on "living". Forty is too fine to maintain, and you end up with categories holding one transaction each.

It knows merchants in five countries

The rules cover chains in the United States, the United Kingdom, Canada, Australia and South Africa. Some examples of what matches without any setup:

Global names — Netflix, Spotify, Amazon, Uber, Airbnb, McDonald's, Starbucks — match everywhere.

Why some rows stay in "Other" on purpose

This is the part most tools do not tell you.

Several large chains share a name with an ordinary English word. Target, Subway, Boots, Three and Metro are all real retailers, and all words that appear in normal transaction descriptions for entirely unrelated reasons — an annual target bonus, a subway season ticket, returned boots.

Matching those would fill in more categories and be wrong some of the time. We leave them out.

That is a deliberate trade, and it runs the way round that respects the reader: a blank category is honest, a wrong one is not. An uncategorised row is visible and takes a second to fix. A row confidently filed under the wrong heading is invisible, and it quietly distorts every total built on top of it.

If your statement is heavy on those merchants, expect a handful of rows in Other and set them yourself.

The transfer problem, which is bigger than the categorising

If you take one thing from this page, take this: money moving between your own accounts is not spending, and no categoriser can tell that it is yours.

A transfer from your current account to your savings account appears on the current account as money leaving. Move it back three weeks later and it appears as money arriving. Do that regularly — a lot of people sweep money to savings and pull some back — and the same money is counted as both income and expenditure, repeatedly.

The effect on the totals is not small. It inflates your apparent income and your apparent spending at the same time, so the two errors hide each other and the picture looks internally consistent while being wrong.

The same applies to credit card payments. Paying a card from a current account is not a purchase; the purchases already happened on the card. Count both and you have double-counted your entire card spend.

Nothing automatic can fix this, because the tool sees a transfer to an account number and has no way to know whose it is. What you can do is deal with it once: identify your own transfers, put them in their own category or delete the rows, and do the same each time you import. It takes a minute and it is the difference between totals you can act on and totals that are simply wrong.

Refunds, reversals and the negative row

A refund arrives as money in. Sorted by amount, it lands with your income.

For a single returned item that is noise. For anyone who buys and returns regularly — clothing especially — it is enough to make a spending category look smaller than it was and a month look better than it went.

If your spreadsheet has a column for the sign of the amount, sort by it and look at everything positive that is not actually income. Most people have three or four rows and are surprised by one of them.

One merchant is not one category

Amazon is the clearest example. The same merchant name covers a book, a washing machine, a grocery delivery and a monthly membership fee — four different categories in the same list, arriving under one description.

A keyword rule has to pick one, and it picks the most common. That is right more often than anything else available, and it is still wrong for you specifically if you use that merchant differently from most people.

The general rule: the categoriser is reliable for merchants that do one thing, and approximate for merchants that do everything. Supermarkets, fuel stations, pharmacies and utilities are the reliable kind. Marketplaces and department stores are not, and are worth a glance.

Two more things that quietly distort a statement

A joint account is two people's spending. Reading joint totals as your own habits is a common and demoralising mistake — you conclude you are worse with money than you are. If the account is shared, the totals describe a household, which is a useful thing to know but a different thing.

Foreign-currency rows carry a conversion you did not choose. A purchase abroad appears in your home currency at the rate and fee your bank applied. That is the correct amount for budgeting — it is what actually left your account — but it means the same purchase can appear at slightly different amounts on different dates, and the bank's fee is baked into the category rather than sitting under Banking & fees.

What to check before trusting the totals

1. Row count. Compare the number of rows to your statement. A converter that drops rows silently is worse than one that fails loudly.
2. The Other bucket. Skim it. If it is large, the statement probably uses unusual merchant descriptors, and a few manual fixes will fix the shape of everything else.
3. Income. Salary lines vary wildly between banks. Confirm your pay landed in Income rather than Other before reading any "you spent X%" figure.
4. Transfers. Look for your own account numbers and card payments in the list, and take them out. This is the one that changes the answer most.

Once those four look right, the totals are worth acting on.

What the categories are for

Sorting is not an end in itself. The reason to do it is that the totals become the actual column in a budget — the thing most budgets never get, which is why most budgets fail in week three.

Read the category totals into the budget planner beside what you planned, and the month stops being a guess. And the merchants that repeat are worth a second pass with the recurring bill detector, which is where the money usually turns out to be.

Try it

Convert and categorise a statement — free for 30 pages a month, no account needed to see the preview.

How does this affect YOUR Money OS?

Every other money decision — a budget, a debt plan, a savings target — rests on knowing what you actually spend. Categorised statements are the cheapest way to stop guessing at that number.

Check my free OS score

FAQ

How do I categorise bank transactions automatically? Convert the statement and let a rule list match each description to a category. Ours does it as part of the conversion, so a PDF comes back as a spreadsheet with every row already sorted.

Is it an AI categoriser? No, and that is deliberate. It is a keyword rule list where the first match wins, which makes it instant, predictable and auditable — the same description always produces the same category.

Why are some transactions uncategorised? Because several large chains share a name with an ordinary English word, and matching them would be wrong some of the time. A blank category is visible and takes a second to fix; a wrong one is invisible and distorts every total above it.

Why do my totals look too high? Almost always transfers. Money moved between your own accounts, and payments to your own credit card, appear as spending — and often as income on the way back. Take them out before reading anything.

Does it work with statements from my country? It recognises chains in the US, UK, Canada, Australia and South Africa, plus global names everywhere. Merchants it does not know land in Other rather than in the wrong place.

Can I trust the category on an Amazon transaction? Approximately. A merchant that sells books, appliances, groceries and a membership under one description cannot be resolved by name alone. Merchants that do one thing — supermarkets, fuel, pharmacies, utilities — are reliable.

Related

• How to convert a bank statement PDF to Excel
• Find every recurring bill and subscription
• Free budget planner