The cash and runway model
This is the method. The inputs are open (N-1), so no figures appear here, but the arithmetic that will consume them is settled and written down. A model nobody can reproduce is an assertion.
Shape
Cash is projected in monthly buckets over an eighteen month horizon. Each month has an opening balance, an inflow, a tax reserve taken off that inflow, an outflow, and a closing balance that becomes the next month's opening. Runway is the number of months before the closing balance first goes below zero, and the zero crossing is reported as a specific month rather than a range.
The projection is stepped, not smoothed. A payment that lands in March lands in March. Averaging it across the quarter produces a prettier line and a worse decision, because it hides the week where the balance actually dips.
Income
Income entries are one of three kinds.
- Certain. Contracted, signed, dated. Counted at face value.
- Recurring. A retainer with a start month and a number of months it runs
for. Counted in each month it covers.
- Probabilistic. A number with a percentage chance of closing attached.
Probabilistic income is the part people get wrong. Counting it at face value produces a plan that assumes every deal lands. Excluding it produces a plan nobody believes, so nobody uses it. We count it three ways and show all three:
| Basis | What counts | What it answers |
|---|---|---|
| Pessimistic | Certain income only | What happens if nothing new closes |
| Expected | Each entry weighted by its probability | The planning number |
| Optimistic | Everything at face value | The ceiling, and only that |
The gap between pessimistic and optimistic is drawn as a band rather than a line, because that gap is the actual state of knowledge and a single line misrepresents it.
Invoices
An invoice contributes cash on the date we expect to be paid, never the date it was sent and never the date the terms say it is due. Those three are different numbers, and the difference between the second and the third is what kills a small business that is technically profitable.
An invoice past its expected pay date is flagged as overdue and keeps contributing on its expected date until that date is revised. Optimism about one invoice should not be invisible; it should be a date someone has to change by hand.
Tax
A percentage of every inflow is fenced off the moment it arrives and is not available to the runway figure. This is the single correction that matters most in a small operation, because money owed to a tax authority sitting in a current account reads as runway and is not.
Without it, every other number in the model is an overstatement.
Expenses
Fixed monthly expenses are itemised rather than entered as one figure, so that a what-if can move one line rather than a total. One-off expenses land on a date and step the line down, which is how a hire or a tax bill actually behaves.
Variable spending is a single monthly estimate. Pretending to forecast it line by line would be false precision.
Scenarios
The three bases above ship as three named scenarios, and they are starting points rather than the ceiling. A scenario can be forked, renamed and kept, so "expected, minus the largest retainer" exists alongside "expected" and both can be put on the same chart.
Nobody models one future. They model the one they fear and the one they are hoping for, and the useful conversation is about the distance between them.
Burn
Burn is reported as the average monthly outflow over the next three modelled months, with a trend against the three after that. It is deliberately not called a trailing average: there is no bank feed behind this model and therefore no history to take a trailing average from. Calling a forward figure a trailing one would be a small lie that makes every derived number harder to trust.
What this model does not do
- It does not connect to a bank, an accounting package or a card.
- It does not forecast revenue. It arithmetic-ally consumes assumptions someone
typed, and the assumptions are the part that is usually wrong.
- It does not calculate tax owed. It reserves a percentage someone chose.
Related methods
Two other methods feed into a full picture and are written up the same way: a subscription view that normalises every billing cycle to a monthly and an annual figure so recurring tooling cost is visible as one number, and a loan schedule that splits a payment into principal and interest and marks the month the two cross over.
The tools implementing all three are being built. They are not linked here until each host verifies live, because a link to something that is not up yet is the kind of thing this desk exists to avoid.