How the budget gets shaped
Most of the false variances in a monthly pack are not caused by the business. They are caused by the budget, specifically by the near-universal practice of taking an annual number and dividing it by twelve. This post is about fixing that without turning budgeting into a project.
Key takeaways
- A flat twelfth is right for very few lines and wrong for most of the interesting ones.
- Three shapes cover almost everything: flat, seasonal, and calendar-fixed.
- Last year’s actuals are the cheapest seasonal shape available and are usually good enough.
- A calendar-fixed line — insurance, an annual licence — goes in the month it is due.
- Reshaping a budget is not rebudgeting: the annual total never changes.
Why a flat twelfth manufactures variances
Take an insurance line budgeted at twelve thousand a year, paid annually in August. A flat twelfth budgets a thousand a month. Eleven months of the year that line is a thousand under budget, and in August it is eleven thousand over. Twelve false variances from a line that was budgeted correctly and paid exactly as expected.
Now do that for the annual licence, the quarterly rates, the seasonal energy bill, the summer wage costs at a seaside business and the January quiet month, and a substantial fraction of your monthly variance table is describing the shape of your budget rather than anything about your business.
Flat, and where it is genuinely right
Rent, monthly subscriptions, a salaried payroll with no seasonal element, a fixed service contract. These are genuinely equal every month and a twelfth is correct. It is worth marking them explicitly as flat rather than leaving them as the default, because a line that is flat by decision behaves differently in the timing filter from one that is flat because nobody thought about it.
Seasonal, from last year
The cheapest useful seasonal shape is last year’s actuals for the same line, expressed as twelve percentages summing to one hundred, applied to this year’s annual budget. It requires no judgement, no forecasting and no meeting. It is wrong in the specific months where last year was unusual, and it is dramatically better than a twelfth for anything with a season in it.
The system can propose these automatically from the prior year’s imported data, which turns budget shaping from a task into a review: here are twelve lines whose shape differs materially from flat, here is last year’s shape, accept or edit.
Calendar-fixed
Insurance in August, the annual licence in March, rates in four instalments. The whole amount goes in the month it is due. This is the single highest-value change to make and it takes about ten minutes for a typical small business, because there are usually only five or six such lines and everybody knows which they are.
Volume-linked lines
The fourth shape is different in kind and worth the extra effort on two or three lines only. A line like packaging or card fees does not have a monthly shape; it has a relationship to a driver, usually sales or units. Budgeting it as a rate rather than an amount means the variance answers a better question.
- Machine learning
- Management
- Analytics
Two or three lines treated this way is worth it; twenty is a modelling exercise that will not be maintained. The candidates are the ones where somebody has said “well, of course it is over, we were busier” more than once.
What reshaping is not
- It is not rebudgeting. The annual total for every line is unchanged, which is why this does not need a board conversation.
- It is not forecasting. Reshaping distributes a number you already agreed; forecasting changes it.
- It is not a one-off. Calendar-fixed dates move, seasons shift, and a fifteen-minute review at the start of each year keeps it honest.
- It is not the system’s decision. The shapes are proposed and a person accepts them, because a budget shape is a statement about the business.
Next: the timing filter, which is where most of the remaining variance goes.
All posts