Comparing months: the seasonality that produces false conclusions

January cost $2,600 and February $2,370. The conclusion suggests itself: spending went down.

The conclusion is almost certainly wrong. February has twenty-eight days rather than thirty-one, so it’s about ten percent shorter by definition. That $2,370 corresponds to roughly $2,620 at January’s length, and nothing was reduced at all.

That’s the simplest of the distortions. Several others exist, and together they account for most of the difference between any two months.

What makes months incomparable

Length. The gap between February and March is three days, around ten percent of everyday spending. Enough to mask a genuine change in behaviour.

Weekend count. A month contains eight weekend days or ten. For most people a weekend day costs more than a weekday, and two extra ones move the total noticeably.

Holidays and time off. A week away restructures the month entirely: transport falls, food rises, categories appear that normally don’t exist.

Shifted charge dates. Rent that landed on the 31st of March and the 1st of May produces an April with two payments and a March with none.

One-off events. Insurance, a repair, a large purchase. A month containing one falls out of the sequence and spoils any comparison it enters.

Making them comparable

Three actions, each taking about a minute.

Calculate per day rather than per month. Divide the total by the number of days and February becomes comparable to March. For everyday categories this is the single most useful normalisation available.

Subtract one-off events. Go through the large transactions and remove anything that happens once a year. What remains is your ordinary month, and the one-offs get compared separately by their own logic, covered in sinking funds.

Compare only controllable categories. Rent and utilities don’t respond to your decisions, their share is large, and they blur the picture. Comparison is meaningful where you have influence.

After those three, the difference between months typically halves, and what remains actually says something about behaviour.

The seasonality nearly everyone has

A year isn’t twelve identical months, and that isn’t a personal quirk.

December costs more almost universally: gifts, gatherings, travel. January runs either expensive through momentum or cheap through long holidays at home, and the spread there is wide.

The start of the school year is its own spike for households with children: supplies, activities, clothes.

A change of season usually means clothes and footwear, and for drivers, tyres too.

Holiday months fall out of the sequence entirely and compare only against other holiday months.

Personal seasonality varies, and it only becomes visible after a year of records. Which is exactly why an annual calendar of large costs is more useful than monthly tables, covered in a financial plan for the year.

What to compare against

The same month last year. The only way to account for seasonality, and it requires a year of data.

A three-month average. The best option before you have a year: it smooths out randomness without going stale.

The previous month. Fine for a quick check, poor for conclusions — too much noise.

Your own plan. Strictly speaking that isn’t comparing months at all, but it’s usually the question you actually want answered: did I stay inside it.

What isn’t worth comparing against: other people’s figures. National averages and a friend’s experience belong to other circumstances, and conclusions from them are generally useless.

How many months a conclusion needs

One month is a data point. Two give a direction without separating a trend from chance. Three are the minimum for a statement about behaviour.

A practical rule: a measure moving one way for three consecutive months is a trend. A measure bouncing up and down is variance, and individual spikes don’t need a response.

Seasonal conclusions need a year, and they only become usable in the second year of tracking, when there’s something to compare against. That’s a long wait, and it’s also when tracking starts answering questions it couldn’t before.

Doing this in the app

The Analytics tab holds a category breakdown for a selected period, and switching periods gives you the basis for comparison.

The app doesn’t normalise per day, so that calculation stays manual: the total divided by the number of days in the month. One division, and without it comparing February to March lies systematically.

A practical format: once a month, write four numbers into a note — the total, the total excluding one-offs, spending per day, and the totals for two or three controllable categories. Three months produces a table showing everything you need, and a year adds seasonality to it.

One check before concluding

Before deciding a month went better or worse, ask yourself one question: did it contain anything the previous one didn’t?

A holiday, illness, visitors, a repair, a job change, a long weekend. If so, comparing it against an ordinary month isn’t valid, and the correct conclusion reads “this month contained that” rather than “I’m spending more.”

The check takes thirty seconds and eliminates most of the false conclusions that lead people to change behaviour in response to a calendar. Telling real growth from apparent growth is covered in why spending grew.