Cash Flow Forecasting: How to Build a Spend Curve on Capital Projects
How to turn a cost baseline into a defensible spend curve: time-phasing against the schedule, S-curve shaping, and the commitment, accrual and cash split.
A £300m capital project gets its baseline approved and the sponsor signs the funding request. Then treasury asks the question the estimate was never built to answer: not how much, but when. How much cash goes out in Q3 next year? What is the peak monthly draw, and in which month does it land? A cost baseline is a total; a cash flow forecast is that total spread across time, and the two are not the same document. Teams that hand finance a lump sum and a completion date force it to guess the profile — and a wrong profile means either idle capital parked in a facility that costs to hold, or a funding gap that stalls the works mid-programme.
This article covers what separates a cost baseline from a cash flow forecast, how to time-phase a baseline against the schedule, how to shape the spend curve so it reflects the way work actually bills rather than a smooth mathematical S, the difference between commitment, accrual, and payment curves, and the failure modes that make a forecast drift. The aim is a spend curve treasury can fund against and a cost engineer can defend in the monthly review — not a straight line drawn between notice-to-proceed and mechanical completion.
Why a cost baseline is not a cash flow forecast
The baseline answers one question well: how much the project should cost at completion. It is a control total, decomposed to a cost breakdown structure, against which every commitment and every accrual is measured. What it does not carry, in its raw form, is time. A £300m baseline is equally consistent with spending £5m in the first month or £25m — the total is silent on the shape.
Cash flow forecasting supplies the missing dimension. It takes each element of the baseline and asks when the money attached to it will actually leave the account, then aggregates those timings into a cumulative curve. The distinction matters because the people who consume the two numbers are different. The project team manages to the total; the finance function manages to the profile. A treasury team funding a portfolio of capital projects needs the monthly draw, the peak, and the confidence around both, because holding cash against a facility has a carrying cost and a funding gap has a delay cost. Give them a total and a date and they will build a straight-line profile by default — which is almost always wrong, because capital projects do not spend linearly.
The consequence of a bad profile is rarely visible until it bites. Under-forecast the ramp and the project draws faster than the facility releases, and procurement slows while approvals catch up. Over-forecast the tail and capital sits committed but unspent, dragging portfolio returns. Neither error shows up in the baseline total — both show up in the cash flow forecast, which is precisely why it is a separate deliverable.
A baseline tells you how much the project will cost; only a time-phased spend curve tells you when the money leaves the account.
Time-phasing the baseline against the schedule
A spend curve is built, not assumed. The mechanically correct way to produce one is to phase each cost element across the duration of the activity that delivers it, using the schedule as the clock. This is why a resource-loaded schedule is the natural home for cash flow forecasting: if every activity carries its cost and its dates, the spend curve falls out of the schedule directly, and it re-phases automatically when the programme moves.
The alignment between the cost structure and the schedule is the prerequisite most teams underestimate. Each control account has to map to the activities that spend it, so that a cost is phased across the right dates rather than smeared across the whole programme. Where the work and cost structures are properly aligned, phasing is a lookup; where they are not, it is a reconciliation exercise every month. For a single account, the spread across its duration is a modelling choice, and three patterns cover most cases:
- Linear. Cost spread evenly across the activity duration. Defensible for steady-rate work such as ongoing site supervision or a stable construction crew, misleading for anything with a ramp.
- Front-loaded. Weighted to the early part of an activity — appropriate for procurement, where a purchase order commits value long before the equipment is delivered or paid.
- Back-loaded or bell-shaped. Weighted late or to the middle — appropriate for commissioning, retention releases, and construction packages that build to a peak before tapering.
Phase every account this way and the aggregate is not a smooth line — it is a curve that starts shallow, steepens through the construction peak, and flattens into the tail. That shape is not decoration; it is the signature of how capital work is actually delivered, and it is the reason a straight-line forecast misleads.
Shaping the curve: why real spend is an S
Aggregate the phased accounts and the cumulative profile of almost every capital project traces an S. Spend is low at the front while engineering and early procurement dominate, accelerates as construction mobilises and multiple work fronts run in parallel, then decelerates through commissioning and closeout. The cumulative curve is the familiar S; the periodic curve — the monthly draw — is a bell that peaks somewhere in the middle third of the programme.
Knowing the shape is useful; knowing where it skews is what makes a forecast defensible. Two projects with the same total and the same duration can carry very different profiles depending on how the work loads.
Reading the skew
- Early-peaking (front-loaded) curves arise where long-lead equipment dominates the cost and commitments land early — common on process and power projects with heavy rotating equipment. The peak arrives before the schedule midpoint.
- Late-peaking (back-loaded) curves arise where construction labour and installation dominate and procurement is comparatively light — common on civil and infrastructure work. The peak sits past the midpoint.
- Symmetric curves are the textbook default and the right starting assumption only when neither procurement nor construction clearly dominates the cost.
A useful discipline is to state the peak monthly draw as a percentage of the total and name the month it lands. On a typical construction-heavy project the peak month often runs in the order of 8–12% of total installed cost, though the figure is entirely project-specific and should be read off the phased model, not assumed. Quoting it forces the forecast to be explicit about the single most important number treasury needs — the maximum monthly exposure — rather than burying it in a cumulative line.
The three curves inside a cash flow forecast: commitment, accrual, and cash
The most common cash flow forecasting error is treating “spend” as a single event. It is not. A pound of project cost passes through three distinct moments, each with its own curve, and confusing them produces a forecast that is internally consistent and practically useless.
Commitment is the point a purchase order or contract is awarded and value is legally obligated. The commitment curve leads — often steeply, because a single equipment PO can commit a large sum in one month long before any cash moves. Accrual is the value of work performed, whether or not it has been invoiced; this is the curve that ties to earned value management, because it measures progress rather than payment. Cash is the actual outflow — invoice paid, net of terms, retention, and milestone conditions — and it lags accrual by the payment period plus whatever is held back.
The gap between accrual and cash is where a lot of forecasts quietly break. Payment terms of 30 to 60 days shift the cash curve a month or two to the right of the work-done curve. Retention — commonly around 5–10% held and released at completion or defects sign-off — carves a slice off the middle of the cash curve and drops it as a lump at the end. Milestone-based contracts step the cash curve rather than sloping it, releasing nothing until each gate is certified. A cash flow forecast that ignores these mechanics will systematically overstate early cash and understate the closeout tail. For treasury the cash curve is the one that matters, but it can only be built correctly by starting from the accrual curve and then applying terms, retention, and milestones — not by relabelling one curve as the other.
Where spend curves drift — and how to hold them
A cash flow forecast is a living forecast, not a one-time deliverable, and four failure patterns recur often enough to be worth naming.
The forecast decouples from the schedule. The programme slips two months but the spend curve still peaks in the original window, so the profile promises cash the project cannot yet absorb. Fix: re-phase the curve every time the schedule is re-baselined or statused — if the forecast is driven off a resource-loaded schedule this is automatic, and if it is not, it is a standing monthly task.
Commitment is mistaken for cash. A large PO is booked and the forecast shows the full value as an outflow that month, overstating the near-term draw by the payment lag and retention. Fix: model the three curves separately and report cash on payment timing, not commitment timing.
The tail is optimistic. Closeout, retention release, final claims, and commissioning support are consistently under-forecast because attention has moved on. Fix: hold an explicit closeout allowance in the curve and resist flattening the tail to zero before defects liability actually ends.
Actuals are never reconciled back. The forecast is issued and never checked against what was drawn, so the same phasing errors repeat every period. Fix: overlay actual spend on the forecast curve each month, and treat a persistent gap between planned and actual draw as a forecasting signal — the same discipline that governs a disciplined monthly cost report applies to the spend curve.
Hold the curve against the schedule, keep the three curves distinct, protect the tail, and reconcile to actuals, and the cash flow forecast stops being a number finance quietly distrusts and becomes the profile the whole capital plan is funded against.
Key Takeaways
- A cost baseline is a total and a cash flow forecast is that total spread across time; the two serve different consumers and are separate deliverables.
- Build the spend curve by phasing each cost account across its activity durations using the schedule as the clock — a resource-loaded schedule makes this automatic.
- Real capital spend traces an S-curve, and stating the peak monthly draw as a percentage of total and the month it lands gives treasury the number it actually needs.
- Cost passes through three distinct curves — commitment, accrual, and cash — and confusing them produces a forecast that is internally consistent but wrong for funding.
- Payment terms, retention, and milestone conditions shift and step the cash curve relative to work done, so the cash profile must be built from the accrual curve, not relabelled from it.
- Spend curves drift when they decouple from the schedule, treat commitments as cash, under-forecast the tail, or are never reconciled to actuals; each has a specific, repeatable fix.