Key takeaways
What this article covers, in order:
- The idea in one example
- The parts of a rule
- Which transactions deserve a rule?
- A worked example: when a rule misfires
- Rules aren't "set and forget"
- Rules versus AI suggestions
What are bank rules? Each one is an instruction you write once: "when a bank transaction looks like this, categorise it as that." Every time a matching transaction comes in, the rule fills in the category for you. Rules are brilliant for regular, predictable payments and quietly dangerous for anything vague, because they do exactly what you told them, even when you told them the wrong thing.
This post explains the idea in general terms, so you understand what you're setting up in whatever bookkeeping tool you use.
The idea in one example
Callum runs a small café in Glasgow. Every Monday a payment goes out to "BELLA DAIRY LTD" for milk and cream. Every week he picks "Food and drink stock" for it. Fifty-two times a year.
A bank rule says: if the description contains "BELLA DAIRY", categorise as Food and drink stock. From then on, Callum doesn't touch it. The rule fills it in, he glances at it, done.
That's all a bank rule is. A condition, and an action.
The parts of a rule
Most rules, in most bookkeeping tools that offer them, are built from the same pieces.
| Part | What it does | Example |
|---|---|---|
| Direction | Applies to money in, money out, or both | Money out only |
| Condition | What to look for in the transaction | Description contains "BELLA DAIRY" |
| Extra conditions | Narrow it down further | Amount is less than £400 |
| Action | What to do when it matches | Categorise as Food and drink stock |
| Optional extras | Add a note, assign a contact, split by percentage | Contact: Bella Dairy Ltd |
Conditions: "contains" versus "equals"
This is where most rules go wrong.
- Contains "TESCO" matches "TESCO STORES 2341", "TESCO PFS" (the petrol station) and "TESCO MOBILE". Three very different costs.
- Equals "TESCO STORES 2341" only matches that exact store, so it misses the branch down the road.
Being too broad gets things wrong. Being too narrow means the rule barely fires. You're looking for the description fragment that's unique to that supplier and that kind of payment.
Amount conditions
Adding an amount range helps tell apart payments to the same name. Callum's dairy order is always between £180 and £320. If a £1,200 payment ever goes to Bella Dairy, it's probably something unusual (a new fridge deal? a backlog of invoices?) and he wants to look at it himself. An amount ceiling makes the rule step aside.
Rule order
When two rules could both match one transaction, the tool has to pick one. Usually that's the first in the list, or the most specific. If you have one rule for "AMAZON" and another for "AMAZON WEB SERVICES", you want the specific one to win. Check how your tool handles order, and put the narrow rules above the broad ones.
Which transactions deserve a rule?
Not everything. A rough test: write a rule when a transaction is regular, the description is consistent, and the category never changes.
Good candidates:
- Rent and service charges to the same landlord
- Direct debits for phone, broadband, insurance, software
- Regular stock suppliers
- Bank charges and interest
- Fuel from the same chain, if you only ever buy fuel there
Poor candidates:
- Supermarkets and big online marketplaces (mixed business and personal)
- Anything starting "PAYPAL " or a card-terminal name like "SQ " or "SUMUP *" followed by a different seller each time
- Transfers between your own accounts, where the reference changes
- Payments that are sometimes capital equipment and sometimes repairs
- Customer receipts, which usually need matching to a specific invoice, not just a category
A worked example: when a rule misfires
Callum sets up a rule: money out, description contains "BOOKER", categorise as Food and drink stock. Booker is his cash-and-carry. Sensible.
In Sep 2026, these went out matching "BOOKER":
| Date | Description | Amount | What it really was |
|---|---|---|---|
| 2 Sep 2026 | BOOKER GLASGOW | £412.80 | Food and drink stock |
| 9 Sep 2026 | BOOKER GLASGOW | £388.15 | Food and drink stock |
| 16 Sep 2026 | BOOKER GLASGOW | £1,140.00 | Commercial coffee grinder (equipment) |
| 23 Sep 2026 | BOOKER GLASGOW | £395.40 | Food and drink stock |
| 30 Sep 2026 | BOOKER GLASGOW | £74.65 | Cleaning products |
Total through the rule: £412.80 + £388.15 + £1,140.00 + £395.40 + £74.65 = £2,411.00, all labelled as stock.
What should have been stock: £412.80 + £388.15 + £395.40 = £1,196.35.
So the rule overstated stock purchases by £2,411.00 − £1,196.35 = £1,214.65. That's the £1,140.00 grinder (which belongs in equipment) plus £74.65 of cleaning products.
His food cost for the month looks like it more than doubled. If he's watching his gross margin, he'd panic. A tweak fixes it: add amount less than £600 to the rule, so the grinder falls out for him to handle by hand. The cleaning products are harder. They come from the same place for similar amounts, and no rule can know what was in the trolley. Callum either accepts a small inaccuracy, splits the line when he spots it, or pays for cleaning products separately.
Rules aren't "set and forget"
Some honest annoyances with rules:
- Descriptions change. A supplier changes bank, or the card processor changes its format, and the rule stops firing. You won't get an error. It just quietly stops.
- Rules don't know about new situations. Start buying equipment from your stock supplier, and the old rule keeps calling it stock.
- Rules multiply. A year in, you might have 80 of them and no idea which one did what.
- Rules can hide mistakes. Because the category is filled in, it's easy to skim past without reading.
A quarterly tidy-up helps: delete rules you no longer need, tighten any that fired on the wrong thing, and check that a few transactions from each rule are right.
Rules versus AI suggestions
Rules and AI categorisation solve the same problem differently.
- A rule is exact and predictable. It does precisely what you wrote, every time, including the wrong thing.
- AI categorisation suggests a category based on patterns. It copes with new suppliers and varied descriptions without you writing anything, but it's a suggestion, not a certainty.
Either way, the safety net is the same: a quick human review of the list, especially anything large or odd, and a monthly reconciliation to prove the books match the bank.
How HelloBooks helps
HelloBooks approaches the "stop categorising the same thing every week" problem with AI rather than rules. On Pro (£14.99 a month), AI auto-categorisation suggests a category for each transaction from your bank feed or CSV import. Everything lands in a review list where you confirm or change it, so a surprise like Callum's grinder is in front of you rather than buried.
The Free plan includes free AI credits to get started, and if credits run out, AI categorisation pauses while the rest of your books keep working. You connect your bank through Open Banking (most UK banks and cards), or import a CSV statement. If you run a café or restaurant, our restaurants page shows how the day-to-day fits together, and the automated bookkeeping page covers the rest.
FAQs
What is a bank rule in bookkeeping?
It's a saved instruction that automatically categorises bank transactions matching conditions you set, such as a word in the description and an amount range.
Are bank rules the same as bank feeds?
No. A bank feed brings transactions into your books. A rule (in tools that have them) decides what category those transactions get.
Why did my rule categorise something wrongly?
Usually the condition is too broad, for example "contains AMAZON", or a one-off purchase happened at a regular supplier. Narrow the description match or add an amount limit.
How many rules should I have?
Only as many as you have truly regular, predictable payments. If you can't remember what a rule is for, you probably don't need it.
Do I still need to check transactions if I use rules or AI?
Yes. A quick review of large or unusual items, and a monthly reconciliation, catch the mistakes automation will always make now and then.
Automation is great at doing the same thing again; your job is spotting the week it shouldn't have.
Start free, no card needed. Try HelloBooks Free
