How to budget on an irregular income (the percentile method)

Nearly every budgeting method you will read assumes something that is not true for you: that the same amount of money arrives on the same day each month.

If you freelance, work on commission, earn tips, contract, or run a seasonal business, the standard advice breaks at the very first step. "Take your monthly income and split it 50/30/20" is unanswerable when the honest reply is which month? And the usual patch — "just use your average" — is worse than no advice at all, for a reason that shows up immediately when you plot the numbers.

This guide is the arithmetic instead. One real-shaped year of freelance income, four calculations, and the reason each one is the calculation to use rather than an alternative.

Why the average is the wrong number

The twelve months above total $33,820, which averages $2,818 a month. That number is arithmetically correct and practically useless.

Look at how many bars clear the dashed line: six out of twelve. Half the year lands below the average, and it lands a long way below — October at $890, April at $640. A budget built on $2,818 is a budget that is wrong in six months out of twelve, and wrong by up to $2,178 in a single month.

This is not bad luck. It is what an average does to a skewed series. One $5,120 month pulls the mean upward for all twelve, so the mean is a number you frequently fail to reach rather than a number you can count on. Averages describe a set of numbers. They do not make a promise about the next one.

What a budget needs is not a description. It needs a floor.

Calculation 1: your self-paycheck

Rule: pay yourself the 25th percentile of your trailing twelve months of net income.

paycheck = PERCENTILE(trailing_12_net, 0.25)

For the year above that is $1,652.50.

The 25th percentile is the value that roughly a quarter of your months fall below — so three months in four, you earn at least this much. Here, nine of the twelve months clear it, which is exactly what you want from a figure you are going to commit to spending every month.

Why not the median? The median ($2,860) is beaten in only six months out of twelve — by definition. A paycheck you can pay yourself half the time is not a paycheck. Why not the minimum ($640)? Because that is the single worst month in a year and living permanently at your worst month is needless austerity; you would end the year with an enormous idle surplus and twelve months of unnecessary restriction.

The 25th percentile sits where those two failures cancel: high enough to be a real income, low enough that you almost always cover it.

Do not compute this from three months of history. With fewer than six data points a percentile is noise wearing a decimal point. Log the months first; the number becomes meaningful once you have half a year and reliable once you have twelve.

Calculation 2: how variable are you, exactly

Two freelancers can both average $2,800 and need completely different systems. The one that matters is dispersion, and the standard measure is the coefficient of variation:

CV = STDEV(trailing_12_net) / AVERAGE(trailing_12_net)

For this year: standard deviation $1,485.98 ÷ average $2,818.33 = CV 0.53.

CV is a ratio, which is what makes it usable — it strips out the size of your income and leaves only the shape, so a $2,000-a-month freelancer and a $20,000-a-month agency can be compared directly. In plain words:

CV	What it means
under 0.20	Steady. Close to a salary.
0.20 – 0.45	Bumpy. Normal for freelance work.
0.45 – 0.75	Volatile. Your buffer has to do real work.
over 0.75	Feast and famine. The buffer is not optional.

At 0.53 this person is squarely in "volatile", which sets up the next number.

Calculation 3: a buffer sized from your own dispersion

"Three to six months of expenses" is the most-repeated emergency-fund rule in personal finance, and for a variable income it is wrong in both directions at once — too much for a steady contractor, dangerously too little for someone at CV 0.8.

Size it from the volatility you actually measured:

months_needed = MIN(12, MAX(3, ROUNDUP(3 + 12 × CV, 0)))
buffer_target = months_needed × (paycheck + fixed_costs)

Three months is the floor for anyone. Every 0.1 of CV adds roughly 1.2 months on top, because a more dispersed income produces both deeper troughs and longer runs of them. Twelve months is the cap — past that you are holding cash that should be working.

With CV 0.53 and fixed costs of $1,850 a month:

• months needed = ROUNDUP(3 + 12 × 0.5273) = ROUNDUP(9.33) = 10 months
• buffer target = 10 × ($1,652.50 + $1,850) = $35,025

That is a big number, and it should be — it is what a CV of 0.53 actually costs to absorb. It is also a target, not an entry requirement. The useful thing about having it computed is that you can see the distance and watch it close, instead of guessing at a round figure someone else picked.

Calculation 4: the number nobody sells you

Here is the one that matters most on a bad Tuesday.

worst_month   = MIN(trailing_12_net)                        = $640
shortfall     = MAX(0, paycheck + fixed_costs − worst_month) = $2,862.50
runway_months = buffer_balance ÷ (paycheck + fixed_costs)

The worst month of this year brought in $640 against $3,502.50 of committed outgoings — a $2,862.50 hole in one month. That is the size of the event your system has to survive, and it already happened once, so it is not hypothetical.

Runway turns your buffer into the only unit that means anything: at zero income, you last N months. A balance of $8,000 is not information. "Four point two months" is.

What actually changes when you do this

Take the self-paycheck every month and leave the rest in the buffer account, and across this year the surplus you hold back is $13,990. The $2,862.50 hole in April is filled by the $3,467.50 that March produced above the line — money that, under the spend-what-lands system, was already gone.

The income did not change. Nothing was earned differently, no client was replaced, no rate was raised. The only change is that a good month is no longer permission to spend and a bad month is no longer an emergency.

That is the whole method, and it is why it survives contact with a real year.

Running it every month

Six steps, once a month, about fifteen minutes:

1. Log every payment as net, after platform fees and business expenses. Gross income is a number for your tax return, not for your household budget.
2. Recompute the paycheck on a rolling twelve months, not on a fixed year. It should drift as your work changes.
3. Move the paycheck to your personal account on the same day each month. One transfer, always the same amount.
4. Move the tax set-aside immediately, as a percentage of gross, into a separate account. This is your money only in the sense that a stranger's coat in your hallway is your coat.
5. Everything left goes to the buffer until it reaches the target, then to whatever you are saving for after that.
6. Recheck the CV every quarter. If it rises, the buffer target rises with it — that is the mechanism working, not a mistake.

The only part that gets tedious is step 2, because a rolling percentile plus a standard deviation plus a conditional buffer rule is not something you want to rebuild in a fresh spreadsheet each month. Set it up once with the formulas above — PERCENTILE, STDEV, AVERAGE, MIN, MAX, ROUNDUP all work identically in Excel, Google Sheets and LibreOffice — and it recalculates itself as you add rows.

The honest limits of this

• It needs history. Under six logged months, none of these numbers mean anything. Start logging now and start using them in the spring.
• It assumes the volatility is roughly stable. A career change or losing your largest client invalidates a trailing twelve months; reset and rebuild.
• It does not create money. If your 25th percentile is below your fixed costs, no budgeting arrangement fixes that — the problem is the income floor or the cost base, and the calculation's job is to tell you so plainly rather than to hide it inside an average.
• Nothing here is tax advice. The set-aside percentage is yours to set or to get from an accountant. The arithmetic just tracks whether you moved the money you said you would.

The short version

Averages describe your past. Percentiles are what you can plan against. Pay yourself the number you beat nine months in twelve, size the buffer from your own measured volatility rather than a stock rule, and know your runway in months rather than in dollars.