Arron Bennett, Founder of Bennett Financials
That's my latest video, on five reasons profitable home service businesses run out of cash. One of them happens every year. Summer is great. Cash comes in fast. So you hire, buy, and take a bigger draw. Then winter shows up, calls drop by half, and the bills don't. This post turns that into a small script that tells you how big your reserve needs to be.
In the video the rule of thumb is two to three months of expenses in the bank at all times. And you fill that up in the busy months before you pay yourself more. The script checks that rule against your actual history.
Full disclosure, I'm bad at math. That's exactly why I like a script doing it.
What does the script measure?
It measures the deepest drawdown. That's the biggest drop in cumulative cash from a high point to a later low point. In a seasonal business, it's how much cash the slow season actually eats before the busy season starts paying it back. If your reserve is smaller than that, you run short every year unless you borrow.
It also works out a simpler floor, two months of costs in your slow months, and uses whichever is bigger.
The data
This is a hypothetical example. The numbers are made up to show the method.
Let's say a 12-person pressure washing and exterior cleaning company does about $1.5 million a year. Most of the work is April to August. Here are two years of monthly cash in and cash out, in thousands of dollars:
import pandas as pd # Hypothetical pressure washing company: two years of monthly cash in and out ($K) months = pd.period_range("2024-11", periods=24, freq="M") cash_in = [62, 48, 41, 55, 98, 182, 205, 200, 190, 162, 104, 79, 66, 50, 44, 58, 103, 190, 215, 210, 198, 170, 109, 83] cash_out = [104, 98, 96, 101, 112, 121, 126, 125, 123, 118, 111, 106, 107, 101, 99, 104, 116, 125, 131, 130, 128, 122, 115, 109] df = pd.DataFrame({"cash_in": cash_in, "cash_out": cash_out}, index=months) df["net"] = df["cash_in"] - df["cash_out"] Use cash, not revenue and expenses from the P&L, because the P&L can look fine while the bank runs dry. Cash in is what customers actually paid that month. Cash out includes payroll, rent, materials, loan payments, owner draws and tax payments. What matters is what happened in the bank.
The calculation
# Deepest run of losses: the biggest drop from a peak in cumulative cash cum = df["net"].cumsum() drawdown = (cum - cum.cummax()).min() # Two months of costs in the slowest months, as a floor slow = df[df["net"] < 0] floor = 2 * slow["cash_out"].mean() reserve = max(-drawdown, floor) busy = df[df["net"] > 0] per_busy_month = reserve / (len(busy) / 2) # busy months per year print(df.tail(12).to_string()) print(f"\nDeepest drawdown: ${-drawdown:,.0f}K") print(f"Two slow months of cost: ${floor:,.0f}K") print(f"Reserve target: ${reserve:,.0f}K") print(f"Set aside per busy month to refill from zero: ${per_busy_month:,.0f}K") cum.cummax() is the highest cumulative cash reached so far. Subtract it from the running total and you get how far below the last peak you are each month. The minimum of that is the deepest hole.
The output
cash_in cash_out net 2025-11 66 107 -41 2025-12 50 101 -51 2026-01 44 99 -55 2026-02 58 104 -46 2026-03 103 116 -13 2026-04 190 125 65 2026-05 215 131 84 2026-06 210 130 80 2026-07 198 128 70 2026-08 170 122 48 2026-09 109 115 -6 2026-10 83 109 -26 Deepest drawdown: $240K Two slow months of cost: $211K Reserve target: $240K Set aside per busy month to refill from zero: $48K What does it tell the owner?
Three things.
First, the business is profitable. Over the last 12 months, cash in beat cash out by about 109 grand. Nothing is broken.
Second, the slow season eats 240 grand. From September to March the business runs short every single month. Seven months in a row. The losses add up to $240,000 before April turns it around. That's more than the two-months-of-costs floor of 211 grand, so the seasonal pattern sets the target, not the rule of thumb.
Third, summer cash isn't summer profit. The five busy months bring in about 347 grand of surplus. To refill a 240 grand reserve from empty, roughly 48 grand of each busy month's surplus has to go into the reserve account. That leaves about 21 grand a month of true extra in the busy season. If the owner takes most of the summer surplus as a draw, winter is guaranteed to hurt.
That's exactly the pattern in the video. Summer feels like the business's real size. The year is the real size.
"Arron, we're up 109 grand on the year. Why are we short every February?" That output is the answer.
How would you use this in practice?
Run it once a year before the busy season, with the latest 24 months of bank data. Set the reserve target from the output. Then automate a transfer. A fixed share of every deposit, from April to August, goes to a separate account until the target is reached. Only after that does the owner's summer bonus come out.
Some easy extensions:
- Add a buffer. Multiply the target by 1.2 to cover a worse-than-usual winter.
- Use three years if you have them. One unusually mild winter can understate the hole.
- Separate tax. Quarterly or annual tax payments can create their own drawdowns. Model them as a separate line so you can see them.
- Check the start month. The drawdown depends on where the history starts. Starting just after a peak is safest.
What are the limits?
It assumes next year looks like the last two. Growth changes the pattern, because more crews mean bigger winter costs. If you're adding people, scale up cash_out for the slow months before you run it. And it's a cash tool, not a profit tool. It tells you how much to hold, not whether your prices are right.
How do you get the data out of your bank?
Most business banks export transactions to CSV. Load it, group by month, and split deposits from withdrawals:
tx = pd.read_csv("bank.csv", parse_dates=["date"]) tx["month"] = tx["date"].dt.to_period("M") monthly = tx.groupby("month")["amount"].agg( cash_in=lambda s: s[s > 0].sum() / 1000, cash_out=lambda s: -s[s < 0].sum() / 1000, ) Clean two things before you trust the result. Remove transfers between your own accounts, or the reserve account will look like spending. And remove draws and repayments on a credit line, because borrowed cash isn't income. If the line covered last winter, the bank data will hide how deep the hole really was. Keep loan payments in cash out only if they're a regular cost, like an equipment loan.
What should the owner do with the number?
Compare it with what's in the bank on the first of April. If the reserve is already at target, the busy-season surplus can go to distributions, equipment or debt. If it's short, the gap tells you exactly how much of the summer has to be held back. So a vague feeling about "saving for winter" turns into one number the owner can check every month.
Why not just use a credit line?
You can bridge a short gap with a line. But a gap you know is coming every year isn't a surprise. It's a cost of the business model, and it should be funded from your own cash. A line used every winter tends to creep up year after year. Interest adds to the cost. And a bank that sees the balance never fall can cut the line in the very month you need it. The reserve does the same job without either risk, and the script tells you exactly how big it needs to be.
For more on the reasoning behind the numbers, I've written about the cash reserve formula for a service business, and our cash reserves research has benchmarks by business type.


