Generate a Cashflow Forecast

This workflow turns a programme and a contract sum into a monthly cashflow forecast: expenditure profile, valuation income, retention, and the net cash position month by month, with an S-curve shape that reflects how the job actually builds. The output is the thing your finance director, your funder and your monthly report all want, and it is the thing most people build by hand in a spreadsheet every month. The point of a cashflow forecast is not the total. You already know the total, it is the contract sum. The point is the trough: the month where you are furthest out of pocket, and how much funding you need to get through it. Everything else in the forecast is scaffolding around that one number. WHEN TO USE THIS Use it at contract award, when you need a first cashflow for the funder or the board and you have a programme but no history. Use it monthly, to reforecast. This is where it earns its keep. The programme has moved, variations have landed, and the profile you drew in month one is now fiction. Rerunning it with current numbers takes minutes. Use it when you are testing a scenario. What does a six-week delay to the frame do to the funding requirement? What if the client's payment terms go from 14 to 30 days? Change one input and look at the trough. Use it on any job where the funding is tight enough that the trough matters, which in UK contracting is most of them. WHEN NOT TO USE THIS Do not use it as your actual cost control. A cashflow forecast is a funding tool. It is not a cost report and it is not a CVR. It tells you when money moves, not whether you are making any. If your margin is wrong, this forecast will show you a beautifully profiled route to losing it on schedule. Do not use AI-generated profiles where you have real data. If you are in month nine of fifteen, your actual valuations to date beat any modelled S-curve. Use the actuals for the months that have happened and only model the remainder. People forget this and re-model the whole job every month, which throws away the best information they have. Do not use it for a genuinely front-loaded or lumpy job without saying so. A scheme that is 60 percent one M&E package delivered in two deliveries does not have an S-curve, it has a staircase. The prompt has a front-end loading field for a reason. Fill it in honestly. Do not hand the raw output to a funder. It is a first draft with modelled numbers in it. Your name goes on the version that goes out, and you should be able to defend every line. COMMON MISTAKES Confusing expenditure with valuation. They are different curves and the gap between them is your problem. Your costs go out when the subbies invoice. Your money comes in after the valuation, after the payment notice, after the payment period. On a monthly valuation with 14-day terms, that is a lag of roughly six weeks between spending and being paid, and if the model does not have that lag in it, the trough will look far shallower than it is. Forgetting retention actually gets held. 3 percent sounds small until you see it as a line on a £14m job. Half of it usually stays put until practical completion and the rest until the end of the defects period, which is often well past the end of your forecast. Model it as held cash, not as a footnote. Trusting the AI's arithmetic. This is the one that catches people. AI models are not calculators. Ask for fifteen months of expenditure that sums to £14.2m and it will produce a table that looks completely right and adds up to £14.34m. Check the totals. Every time. If the cumulative column does not land exactly on the contract sum, the whole forecast is wrong and it is wrong in a way that looks fine. FREQUENTLY ASKED QUESTIONS What is an S-curve and why does construction spend follow one? The cumulative expenditure curve on most construction projects is S-shaped: slow at the start, steep in the middle, flattening at the end. It happens because of how jobs mobilise. Early on you have a small gang doing enabling and substructure. In the middle you have the frame, envelope and M&E rough-in all running at once with maximum labour on site. At the end you are down to snagging and commissioning with a handful of people. Draw cumulative spend against time and you get an S. If your forecast is a straight line, it is not a forecast, it is a division sum. How accurate is an AI-generated cashflow? The shape is usually reasonable. The numbers are as good as your inputs and the arithmetic needs checking. Treat it as a well-informed first draft that saves you an hour of spreadsheet work, not as an output you can issue. The one thing it genuinely cannot know is your job: that the steel is on a six-week lead and lands in one payment, or that the client always pays on day 29. Should the forecast be based on cost or on value? Both, and that is the entire point of the exercise. Model expenditure (what you pay out) and valuation income (what you get in) as separate lines, then the net position is the difference. A forecast with only one of them tells you nothing about funding. If you are only being asked for one, ask which question they are actually trying to answer. How does retention affect the cashflow? It withholds a percentage of every valuation, commonly 3 to 5 percent on UK contracts, usually with half released at practical completion and the balance at the end of the defects liability period. On a £14m job at 3 percent that is £426,000 sitting with the client, and the second half of it typically falls outside a 15-month forecast entirely. Model it explicitly as a held line, because it is real money and it is the difference between the forecast and the bank. What payment terms should I model? Whatever your contract says, not what you hope. Under the Housing Grants, Construction and Regeneration Act 1996 as amended, the contract sets the due date and the final date for payment, and the payment notice and pay less notice regime hangs off those. Read the dates out of your contract and model those. And model your subcontract payment terms separately, because the gap between when you pay them and when you get paid is exactly what the forecast is measuring.

Use at contract award to build the first cashflow, then monthly to reforecast as the programme moves and variations land. Input the contract value, the dates, the retention and payment terms, and a short description of how the spend actually profiles, and you get a monthly table with expenditure, valuation income, retention held and the net cash position, plus the funding trough and an S-curve description. Also useful for scenario testing before a commercial meeting. Change the payment terms or push a key package back six weeks and see what happens to the funding requirement. One rule: once the job is running, feed in your actual valuations for the months that have happened and only model the remainder. Real data beats a modelled curve every time.