The categories you pick decide what a budget can tell you

This looks at spending categories from the practical end — what holds up once conditions stop being ideal.

A category is a question, not a label

Every category you create commits you to answering one question each month, and categories that answer nothing are pure administration. The useful test is whether a number could change a decision, since knowing the transport total only helps if transport is negotiable.

Categories that mirror the shops you use rather than the choices you make produce data that is accurate and entirely inert. Grouping by decision means combining anything you would change together and separating anything you would change independently. That is why food eaten at home and food bought out usually belong apart, while three streaming services usually belong together.

Too many and too few

Very granular schemes collapse within two months because classifying a mixed supermarket receipt takes longer than the resulting insight is worth. Very coarse schemes survive but reveal nothing, since a single large category can hide an entire change in behaviour inside it. Most households function on somewhere between eight and fifteen categories, though the right number depends on how much spending is discretionary.

A practical approach is to start coarse and then split only the category that has repeatedly surprised you. Splitting in response to a surprise keeps the scheme small and ensures every subdivision exists for a reason you can still remember.

The miscellaneous problem

A large other category is the clearest sign that the scheme does not match the household, because it absorbs whatever the design failed to anticipate. The fix is not to abolish it but to read it, since the items appearing there repeatedly are usually a missing category.

A small residual bucket is healthy and stops the whole system stalling on a transaction that genuinely fits nowhere. Where the residual grows past a modest share of variable spending, the budget has stopped describing what is actually happening. Reviewing that bucket once a quarter is usually enough to keep the scheme accurate without constant redesign.

Why automatic categorisation misleads

Software assigns categories from merchant codes and names, which describe who was paid rather than what was bought. A supermarket that also sells fuel, clothing and household goods appears as groceries regardless of what was in the basket.

General retailers and online marketplaces are worse, because a single merchant can legitimately map to half a dozen categories. The result is a report that is precise, internally consistent and wrong in exactly the places that matter most.

Correcting a handful of large transactions each month recovers most of the accuracy without reviewing every line.

Fixed, committed and variable

Alongside the topical categories, tagging each cost as fixed, committed or genuinely variable answers a different and more important question. Fixed costs cannot change this month, committed costs can change with notice, and variable costs are decided transaction by transaction. That split shows how much of the budget is actually available if income falls, which the topical categories cannot reveal.

Practically, households often discover that the proportion they can change quickly is far smaller than they had assumed. The same figure is the one that matters when judging whether a new subscription or agreement is affordable at all.

The right answer depends on your tax situation, which this cannot see.

Changing the scheme without losing the history

Renaming or merging categories mid-year breaks comparison, which is the main thing a second year of data is for. Where a change is necessary, keeping the old category as a closed heading rather than deleting it preserves the earlier months.

On the balance sheet, making structural changes at the start of a year, and only once, is enough for almost every household. The exception is a genuine change in circumstances, since a new job, a child or a move invalidates the previous scheme anyway. A budget redesigned constantly produces twelve unconnected months rather than one comparable year.

The takeaway

Build categories around decisions you could make, not shops you visit, and treat a growing miscellaneous bucket as a design fault.

Costs compound as reliably as returns do, and in the same direction.