How to build a cash flow forecast step by step

Learning how to do a cash flow forecast is less about accounting and more about running the business with your eyes open. If you own or manage a company of ten, fifty or a hundred and fifty people, what you need on Monday morning is a clear answer to one question: how much cash will I have in six weeks, and in six months, if nothing surprising happens? If you cannot answer that, every decision about hiring, buying equipment or weathering a slow quarter is taken half blind.

The idea is simple. You start from the real cash sitting in your bank accounts today, add what you will actually collect on the dates you will actually collect it, subtract what you will pay on the dates you will pay it, and watch the closing balance roll forward month by month. The craft is in not making up the collection dates, not forgetting the quarterly tax bills, and reviewing the forecast every week so it keeps up with reality.

This guide walks you through it step by step, with a fictional engineering SME of twenty-four people and 1.8 million euros of annual turnover. At the end you have the full six-month table and the mistakes you are most likely to make the first time.

What a cash flow forecast includes (and what it does not)

Anything that moves money in or out of your bank accounts belongs in the forecast: customer collections, supplier payments, payroll, rent, loan instalments, taxes, insurance, capital expenditure and one-offs. The profit and loss statement does not. An invoice you raise in November and collect in February is February cash, not November cash. That one distinction explains why a profitable business can run out of money. Depreciation, provisions and reclassifications stay out of the forecast too, because they never hit the bank. If you want a longer take, the piece on what cash flow actually means for a business takes its time.

Step 1: pick the horizon and the level of detail

The right horizon depends on what you want the forecast for. For day-to-day operational decisions, thirteen weeks with weekly detail is the standard: enough to see a full quarter and to spot a dip before it hits. For hiring, investment or a conversation with the bank, six to twelve months with monthly detail.

Most small finance teams run both views at once. The weekly thirteen weeks for the near term, the monthly view for the horizon. The closer the time, the finer the grain; the further out, the coarser. If you want the mechanics of the near-term view, the dedicated guide to the thirteen-week forecast has the format and the cut-off rules.

Step 2: start from the real bank balance, not the ledger

The first figure in your forecast is today’s cash balance across all your current accounts, summed. That number comes from the bank, not the books. If you run several accounts, several entities or a credit line with partial drawdown, write one row per account and total at the bottom.

A classic trap: the bank balance and the ledger balance rarely match. Cheques issued but not presented, deposits in transit, unposted entries from yesterday, or old reconciling items dragging along. Before you build anything, reconcile and anchor the forecast to the real bank figure. The gap between bank balance and book balance in Odoo deserves its own chapter, and it is one of the most common reasons a forecast looks off when it is not actually failing.

Step 3: schedule collections by the real pay date, not the invoice date

This is where half the game is won. Invoice date and real pay date almost never agree. If you sell on thirty-day terms to clients who in reality pay you at sixty, seventy-five or ninety days, dropping each invoice into its invoice month gives you a tidy forecast that will not hold up.

What makes the forecast useful is working out, client by client or by cluster of clients, how long each one actually takes to pay. If you have four large clients and twenty smaller ones, look at the last twelve months and compute the real average delay of the four large ones individually; cluster the rest with a shared average. Then apply that delay to every open invoice and drop the amount in the month you expect it to land.

For future sales not yet invoiced, use the planned invoicing schedule plus the same delay. For overdue invoices, decide case by case and err prudently. Two numbers stabilise the model: your DSO (Days Sales Outstanding, how many days a euro of sales takes to turn into cash) and the historical bad debt rate (if 2% never comes in, apply the haircut as a buffer). When late payment is a defining feature of the business, cash scenarios for small businesses when a client pays late show how to layer a “key client slips by 60 days” scenario on top of the base plan without rewriting it.

Step 4: list every payment, one by one

Payments are easier than collections, because you control most of them. But a payment or two always goes missing on the first pass. The minimum list looks like this:

Payroll and social contributions. Payroll on your usual pay date, employer contributions and PAYE (or the local equivalent) a few weeks after. Add the year-end bonus to its month. Severance and variable pay belong on their own line.
Suppliers. Enter each supplier invoice in the month of its due date, not the month it was issued. Keep a short list of recurring payments (rent, utilities, insurance, SaaS, maintenance) that you drag onto every month.
Taxes. The VAT return is the single most forgotten payment. In the UK it is usually quarterly; in Ireland, usually every two months, by the 19th (the 23rd if you file and pay through ROS). Add corporation tax instalments, payroll taxes and anything local that applies to you. Timing varies by country, which means a chunky payment lands several times a year. Tax is covered in detail in how to handle taxes in a cash flow forecast.
Loans and leases. The monthly instalment is known; drag the full line to the last payment of the schedule.
Annuals, one-offs and distributions. Public liability, accountancy fees, software licences, agreed dividends: whatever gets paid once a year quietly disappears from spreadsheets. Keep them in.

Step 5: capital expenditure and the minimum safety balance

Two things close out the model. First, one-offs: a new machine, a fit-out, a legal settlement, an acquisition payment. Anything outside the normal rhythm goes on its own row in its month, so it does not blur into the operating pattern.

Second, the minimum safety balance. Decide the floor you never want to cross. For a healthy SME that is usually one month of operating payments (payroll, suppliers, rent, loan instalments); one and a half or two months if your collection cycle is long. It is the red line of the forecast: if any month’s closing balance falls below it, something needs to move before you get there.

Worked example: Weald Engineering Ltd, six months, month by month

Weald Engineering Ltd is a small industrial maintenance engineering firm, twenty-four people, 1.8 million euros of annual sales, four large clients that in practice pay at 55 days and around twenty smaller ones at 35. Billing is fairly steady, with the usual January dip. Opening bank balance on 1 November 2026: €68,400. One full extra month of payroll in December (bonus). Quarterly VAT and corporation tax instalments on dates set by the company’s tax calendar. A new inspection rig planned for March at €18,000. Minimum safety balance set by the owner: €60,000. Adjust the months to your own tax calendar.

Line	Nov	Dec	Jan	Feb	Mar	Apr
Opening balance	68,400	112,400	100,400	90,900	137,900	166,900
Customer collections	155,000	145,000	110,000	165,000	170,000	150,000
Payroll and social	52,000	52,000	52,000	52,000	52,000	52,000
December bonus	0	52,000	0	0	0	0
Suppliers	48,000	42,000	38,000	55,000	60,000	50,000
Rent and insurance	6,800	6,800	6,800	6,800	6,800	6,800
Loan instalment	4,200	4,200	4,200	4,200	4,200	4,200
Quarterly VAT	0	0	18,500	0	0	16,800
Corporation tax instalment	0	0	0	0	0	9,600
New equipment	0	0	0	0	18,000	0
Closing balance	112,400	100,400	90,900	137,900	166,900	177,500

The low point sits at €90,900 in January, well above the €60,000 floor the business has set. The point of the example is not the numbers, it is that January’s dip is not a surprise. It is visible in November, with two months to react: pull a large collection forward, defer a non-essential purchase, draw on a short-term facility if needed. That is the reason the forecast exists.

Step 6: review weekly, planned against actual

A forecast that never gets reviewed stops working within three weeks. Set a fixed slot (Monday morning works), take last week’s bank statement and compare it to what you had planned. The usual outcomes:

A big collection has slipped a week. Move the line and check the floor is still safe.
A client paid earlier than expected. Add the margin to next month.
A payment appeared that you had not forecast (a refund, a social security adjustment, a tax reassessment). Add it as a permanent line if it can repeat.
Payroll or supplier payments came in at a slightly different amount. Adjust the pattern.

The value is in writing down why the numbers moved, not just in matching them. If every week you jot two lines “collection X delayed because of Y”, after three months you have the real map of your collection cycle, and the forecast tightens without any effort.

The mistakes most SMEs make the first time

Seven or eight slip-ups come up in almost every first pass:

Scheduling collections by invoice date instead of real pay date. The forecast looks neat and misses every month.
Forgetting quarterly taxes. The VAT bill in January is invisible until it lands.
Leaving out the bonus month. December is the biggest forecast-breaker.
Anchoring on the ledger balance, not the bank. If the banks are not reconciled, the error travels forward from column one.
Mixing group companies. If you run two entities that invoice each other, posting the intercompany flows on both sides inflates sales and purchases and hides the real position. Intercompany belongs on its own line and gets netted out.
Not updating the collection pattern when a client’s behaviour changes. If a key client has moved from 45 to 75 days for three months running, update the assumption.
Chasing precision to the cent. It is an estimate. Round to the nearest hundred or thousand euros; it is enough to decide.
Not reviewing. Building the forecast once a month and ignoring it until the next close is almost the same as not having one.

When to leave the spreadsheet behind

Spreadsheets are fine while the business fits on one tab. Once you have several bank accounts, two or three legal entities, hundreds of invoices a month and a handful of clients whose timing moves the monthly balance, the spreadsheet gets brittle. Every month you paste statements, update formulas, pray no one has edited a row and get the picture a week late. In the honest view of forecasting in Excel there is a list of what the spreadsheet does well and what it stops doing above a certain size.

Beyond that size, it pays to look at tools that read the data you already have. If you run Odoo, the obvious move is a tool that reads your sales and purchase invoices with due dates, your payments, your reconciled bank statements, your loan schedules and your group entities, and projects on top. Building the forecast on Odoo has its nuances between Community and Enterprise, but it is the short route when the data already lives there.

Kashbi is the piece we have built for that case: it reads your Odoo (one or several companies), shows the real cash of every account up to date, projects collections and payments by their due dates, pulls in loan instalments and the estimated VAT and withholdings, and emails you before the balance crosses the floor you set. It opens soon.

Frequently asked questions

How far out should a small business forecast cash?

Thirteen weeks with weekly detail for daily management, and six to twelve months with monthly detail for hiring, investment or bank conversations. Running both at once is normal; the near-term one gets a weekly review, the longer one gets refreshed at month end.

How do I include VAT if I do not know the quarterly figure yet?

Estimate output VAT from the sales forecast and input VAT from the purchase forecast, and project the net payable or refundable. For the first version, use the average rate from the last four quarters. Then every month, tighten the estimate with real figures.

Should I use the invoice date or the expected pay date?

Always the expected pay date. The invoice date is for accounting and for measuring sales, not for knowing how much cash you will have. If your biggest clients routinely pay at 60 or 75 days, modelling them at 30 draws a cash position that does not exist.

How often should the forecast be reviewed?

Weekly for the thirteen-week view, so you can shift collections and payments as reality lands. Monthly for the six to twelve month view, at close. The expensive part is building it once and then dropping the habit.