Cash Flow Forecasting: What Goes Wrong on Construction Projects

The Forecast Is Built on Assumptions That Die on Day One

Every cash flow forecast starts with a set of assumptions. The programme will run to schedule. Payment claims will be submitted monthly on the contractual dates. Variations will be assessed and valued promptly. Retention will be held and released per the standard terms. The contractor will claim in line with the priced schedule of quantities.

On a $60M infrastructure project, the original forecast might show a smooth S-curve of expenditure over 18 months. It looks clean. It makes sense. The QS signed off on it. The board approved it.

Then construction starts.

The earthworks take three weeks longer than planned because of weather. The contractor front-loads their first payment claim to improve early cash flow. A design coordination issue triggers two variations totalling $400K that were not in the original model. The programme slips by five weeks but the forecast still shows the original completion date.

A cash flow forecast that is not connected to the live programme, the approved variations register, and the actual payment claim history is not a forecast. It is a guess wearing a spreadsheet.

The Five Things That Break Every Forecast

After 10 years managing projects from $10M to $750M, I have seen the same five problems kill cash flow accuracy on almost every project.

1. Programme slippage without forecast adjustment

The contractor's programme under Clause 9.2.1 of NZS 3910 is the backbone of the cash flow forecast. When the programme slips, the expenditure profile shifts. But on most projects, the forecast is not updated to reflect programme changes. The result is a forecast that shows money being spent in months when the work has not yet started.

2. Variations approved but not modelled

A variation under NZS 3910 can take weeks or months to be formally valued and approved. During that time, the work is being done and the cost is being incurred, but the forecast does not reflect it. When the variation is finally approved at $250K, the forecast suddenly jumps. The board asks why. Nobody has a good answer because the variation was known about for months but never included in the cash flow model.

3. Retention treatment is wrong

Retention under Clause 12.4 is straightforward in the standard form. Half released at Practical Completion, half at the end of the Defects Liability Period. But Special Conditions regularly amend these terms. Some contracts reduce the retention percentage. Some change the release triggers. Some introduce performance-based retention. If the forecast uses the standard form assumptions when the Special Conditions say something different, the cash flow is wrong from the first claim.

4. Payment claim timing does not match reality

The forecast assumes monthly payment claims on set dates. In practice, claims are submitted late, disputed, or adjusted. Under the Construction Contracts Act, payment schedules have strict response deadlines. If a claim is larger than expected, or contains disputed items, the actual cash outflow can differ significantly from what the forecast shows. One large disputed claim can throw the quarterly cash position out by hundreds of thousands of dollars.

5. Nobody owns the forecast after it is created

This is the root cause. The QS builds the forecast during preconstruction. It gets approved. Then it sits in a folder. Nobody updates it. Nobody reconciles it against actual claims. Nobody adjusts it when the programme changes. The forecast becomes a historical document rather than a management tool.

The problem with most cash flow forecasts is not the model. It is the absence of a process to keep the model connected to reality.

What a Good Forecast Looks Like

A forecast that works is not more complex than a forecast that fails. It just stays connected to the project.

• Tied to the current programme. When the programme is re-baselined, the forecast updates. The expenditure profile follows the work, not the original plan.
• Includes approved and pending variations. Approved variations are in the base forecast. Pending variations are modelled as probability-weighted items so the range of likely outturn cost is visible.
• Retention modelled on actual contract terms. Not the standard form. The actual Special Conditions that apply to this contract, including any amendments to percentage, release triggers, or bond substitution options.
• Reconciled monthly against actual claims. Every month, the forecast is compared to the actual payment claim. The variance is explained. The forecast is adjusted. Over time, the accuracy improves because the model learns from reality.
• Owned by someone. One person is responsible for keeping the forecast current. It is not a shared responsibility. It is not "the QS's job" if the QS left the project six months ago.

The Forecast as an Early Warning System

A good forecast does more than tell you where the money is going. It tells you where trouble is building.

If the forecast shows the project trending 8% above the approved budget and nobody has flagged it, that is a governance failure. If the forecast shows three months of under-claiming followed by a large catch-up claim, that usually signals the contractor is under financial pressure. If the forecast shows retention release dates approaching but the defects register has not been reviewed, you are about to release money for work that may not be complete.

A cash flow forecast should be the first place a project director looks when they want to understand the financial health of a project. If they cannot trust it, they are flying blind on a multi-million dollar commitment.

The forecast is connected to everything else on the project. The cost overrun risk, the programme status, the variation register, the payment claim history. When those inputs are tracked properly and fed into the forecast automatically, the model stays accurate. When they are not, the forecast becomes a document nobody trusts and nobody uses.

Stop Forecasting Once and Start Forecasting Continuously

The shift is simple but requires discipline. Stop treating the cash flow forecast as a document produced at the start of the project. Start treating it as a living model that is updated every time something changes. Every approved variation. Every programme revision. Every payment claim.

This is not additional work. It is the same work your team is already doing. The difference is connecting the dots. When the variation register updates, the forecast updates. When the programme re-baselines, the expenditure profile shifts. When a claim is assessed, the actuals feed back into the model.

The firms that do this well do not get surprised. They know where the money is, where it is going, and where the risks sit. They walk into board meetings with confidence because the numbers are real. That is what good forecasting looks like.