Three months in and my reports are meaningless because my categories have turned to mush, how do people keep them usable?

Ninety categories is the standard failure and the fix is the opposite of what feels rigorous: use about eight, and make each one a decision you could actually make.

The test for whether a category earns its place: would you ever change your behaviour based on that number alone? If the answer is no, it is detail, not a category.

Nobody has ever changed anything because household was four pounds higher than last month. People do change things when they see what eating out or transport costs over a quarter.

A structure that survives, roughly:

• Fixed, rent, bills, insurance, subscriptions. Things you cannot change this month.
• Food in, everything eaten at home, including the supermarket, including the corner shop.
• Food out: restaurants, takeaway, coffee.
• Transport
• Personal, clothes, hobbies, haircuts, whatever
• Health
• Gifts and one-offs
• Everything else

Eight lines you can read in ten seconds.

The two that matter most are food in versus food out, because that split is genuinely actionable and it is the one people most often blur into a single food category. Your groceries, supermarket, food and household confusion resolves into one line: food in.

On stopping the drift, since the structure only survives if adding a category is hard.

One rule: never create a category to file a single transaction. That single rule prevents ninety. If something does not fit, it goes in everything else, and if everything else becomes large and interesting, then you split it: deliberately, once, with a reason.

Write the rule for each category somewhere. One line each: food in means anything bought to eat at home, including the corner shop. It sounds excessive and it is what stops you making a different decision in March than you made in January. Your groceries-versus-supermarket problem is entirely a missing definition.

Let the automatic rules do the boring majority. Most apps will learn or let you set a rule per merchant. Set those for your twenty most frequent merchants and ninety percent of transactions file themselves correctly forever.

Hand-sort the remainder once a week, in one sitting. Five minutes.

Split transactions sparingly. Splitting a supermarket shop into food and household is where a lot of effort goes for a number nobody acts on. Put the whole thing in food in and accept a small imprecision.

Precision is not the goal. A number you look at is worth more than a precise one you do not.

Groceries and supermarket and food as three separate categories is the exact moment it stopped being useful.

The part your question is circling and has not quite reached: recording is not the habit. Reviewing is.

Three months of faithful recording that you have never acted on is three months of admin. The categories only matter because they feed a decision, and if there is no regular moment where a decision gets made, no structure will help.

So build the review before you rebuild the categories:

A fixed slot, weekly, ten minutes. Same time every week. Look at the eight numbers, ask whether anything surprises you, and if something does, decide one thing to change.

One monthly comparison against the previous month, and that is all.

No more than one change at a time. The point is not to optimise everything; it is that the numbers cause something to happen.

On starting again: do not re-categorise three months of history. It is a large amount of work for a report you have already decided is meaningless, and the temptation to do it is what stops people restarting at all. Draw a line, set up eight categories from today, and let the old data be old data.

If you want a baseline, spend twenty minutes roughly totalling the old months into your new eight. Approximate is fine. It is a comparison point, not accounting.