Bank Rules That Actually Work (And the Ones That Backfire)

Bank rules are the cheapest automation in QuickBooks Online and the one most likely to be doing damage right now without anyone noticing. They run silently, they run on every transaction, and when they are wrong they are wrong consistently.

We open a lot of files where the rule list has grown to sixty entries nobody remembers writing. Half overlap. Two of them fight. Here is how to build a set that holds up.

Match on the payee string, not the amount

Amount-based rules feel precise and are the most common way people poison a file. The vendor raises their price, the rule stops matching, and the transaction quietly falls back to whatever guess QuickBooks makes. Nobody notices until the quarter looks weird.

Scope on the bank description instead, and scope it tightly. Use contains with the longest stable fragment of the merchant string — not "AMZN" but the full processor prefix your bank actually sends. Read the raw feed text before writing the rule, not your memory of it.

Auto-add only what is truly deterministic

Auto-add is the difference between a rule that suggests and a rule that acts. Reserve it for transactions where payee, account, and treatment are identical every time: the SaaS subscription, the processor fee. Everything else stays a suggestion a human confirms. If a rule requires judgment — which class, which customer, whether this one is a refund — it does not get auto-add. Ever.

A rule that acts on your behalf must be one you would bet the quarter on. If you would not, leave it as a suggestion.

Order matters more than people think

Rules evaluate top to bottom and the first match wins, so specific rules must sit above general ones. If a broad "contains AMEX" rule sits at the top, the twelve careful rules underneath will never fire.

Write specific first, general last, and keep a catch-all at the bottom that routes to a review account rather than something plausible. An obviously wrong bucket gets fixed. A plausible one does not.

Rules cannot rescue a bad chart of accounts

Most rule chaos is downstream of account chaos. When three accounts all mean software, people write three rules and the split is arbitrary forever. Fix the account structure, then write rules against it. Rules encode decisions; they do not make them.

Audit the list every quarter

Sort by last-modified and read the list end to end. Anything that has not fired in ninety days is dead weight from a vendor you dropped. Delete it. Anything auto-adding to an account you cannot immediately explain gets demoted to a suggestion until someone can. Fifteen rules you understand beat sixty you inherited.

Want the feed and rule set built properly the first time? That is Bank & App Connections, and the messier categorization logic lands in Workflow Automation.