The Project Is Finished. The Costs Aren’t.
Why modern project cost control must look beyond actuals and commitments
A completion notice is not a cost forecast
The bill of materials is closed. Manufacturing has reported completion. The project dashboard is reassuringly green. Yet somebody on site is waiting for a missing component, a subcontractor has not sent the final invoice, and the assembly team has already spent hours that nobody has included in the remaining-cost estimate. The project may be complete in the system. Economically, it is not.
This is one of the less glamorous problems in industrial project business. It rarely produces a dramatic warning. Instead, it arrives as a succession of perfectly reasonable explanations: the invoice came late; the work order had been closed; the purchase order was already settled; the remaining installation work was considered insignificant. Each explanation may be correct. Together they can produce a seriously misleading forecast.
The distinction matters. An accounting report answers what has been recorded. A project forecast must answer what the project will ultimately cost. Those are related questions, but they are not the same question. Confusing them is an expensive habit.
The costs that fall between the categories
In a typical project review, three figures dominate the discussion: the original estimate, actual costs and outstanding commitments. The problem begins when managers assume that these three figures describe the entire economic position. They do not. Work may have been performed but not invoiced. Work may be unavoidable but not yet ordered. A technical completion notice may have been entered while corrective work remains open.
I distinguish three practical sources of trailing costs. First, costs already incurred but not yet recorded: unbilled subcontractor services, time sheets awaiting processing or transport charges still in transit. Second, work still to be performed although the relevant task or bill of materials has been declared complete: repairs, missing parts, tests or additional assembly hours. Third, consequences of unresolved interfaces: a delivery arrives on time but cannot be installed, or a design change shifts effort from engineering into the workshop and then onto the site. The distinction becomes visible in a familiar sequence. Engineering releases a bill of materials, purchasing places the orders, and manufacturing reports its package complete. A missing bracket is then identified during assembly. It requires a drawing clarification, a small additional order, transport and several hours of fitting. None of those consequences necessarily appears in the closed manufacturing package when the completion message is issued. The bracket is inexpensive; the interruption it causes may not be. The controlling question is therefore not whether the original package was correctly closed in the system, but whether the full chain of work needed to finish the deliverable has been recognised economically.
These are not three new accounting categories. They are three questions that the project team must ask before accepting the cost forecast. A forecast that ignores them is not necessarily dishonest. It is incomplete. And incompleteness is particularly difficult to spot when every individual report appears accurate.
The difficult cases are rarely hidden in a single large invoice. They are scattered across the boundaries of work packages. A supplier has delivered the machine but not the final documentation. Manufacturing has released the assembly but a corrective operation is still pending. Installation is nominally complete while the customer is waiting for a successful test. Each function may have a defensible reason for calling its own task finished. None of those statements proves that the project has no further cost exposure. The project manager must reconcile the different definitions of completion before accepting the financial picture. Interfaces are particularly vulnerable. Manufacturing may regard a component as delivered, while assembly cannot use it without modification. Engineering may regard a drawing revision as completed, while purchasing has not yet identified the changed material requirement. Site management may know that a subcontractor is staying longer, but the commercial consequence has not reached the project forecast. Each department can describe its own status accurately, and the overall forecast can still be wrong. A reliable remaining-cost discussion must therefore cross the handover points rather than merely add up departmental reports.

Figure 1. The hidden cost gap between system records and the work still to be recognised.
A remaining-cost bridge, not another dashboard
The practical answer is a disciplined remaining-cost bridge. Start with posted actuals. Add open commitments, taking care to understand what those commitments cover. Then identify work still required but not yet committed. Check separately for costs already incurred but not yet posted. Finally, assess risks that could affect the final cost. The categories must not overlap. Counting the same subcontractor invoice as an open commitment, an unposted cost and a risk would make the forecast just as unreliable as omitting it altogether.
This sounds straightforward. In real projects it demands technical knowledge. Somebody must know whether the outstanding purchase order includes commissioning. Somebody must check whether a supposedly completed work package still needs rework. Somebody must distinguish a contractual claim that may be recovered from a cost that the project will incur regardless of whether the claim succeeds.
The useful unit of discussion is not a colourful traffic light. It is a named issue: what changed, which quantity or work package is affected, what it may cost, who owns the assessment and when the forecast will be updated. Where action is possible, the issue must also carry a decision. Otherwise the project team has merely described the problem more elegantly.
The bridge should follow the logic of the original estimate. If quantities were estimated by equipment group, discipline or installation package, the project should be able to trace material changes through those same units. When the estimate and the execution report use entirely different structures, the organisation spends the rest of the project translating its own numbers.
I would also distinguish between a missing cost and an unresolved commercial claim. The project may be entitled to recover the cost of a customer-driven change, but that entitlement does not make the cost disappear. Cost to complete and expected recovery must be evaluated separately, with an explicit assessment of the contractual position. Otherwise an optimistic claim quietly offsets a very real expense. A project forecast should not depend on winning an argument that has not yet taken place. A commercially responsible forecast shows both the gross exposure and the possible recovery, so management can judge the risk without confusing a negotiation position with earned revenue. This separation also makes the next decision clearer: secure the evidence, agree the contractual response and continue controlling the work itself.
The first warning is usually a conversation
The earliest reliable warning seldom comes from the month-end close. It comes from an engineer who notices a design assumption no longer holds, a foreman who says that the planned assembly sequence will not work, or a buyer who discovers that the cheap component requires expensive modifications elsewhere. None of these observations is automatically a cost forecast. But each is a reason to ask whether the forecast remains credible.
That is why I attach considerable importance to Controlling by Walking Around. Not as a management fashion, and certainly not as a ceremonial tour of the shop floor. The controller or commercial project manager must go where the work is happening, ask sensible questions and listen carefully. What is still open? What was marked complete prematurely? Which tasks are waiting for parts? Which additional hours are already foreseeable? What do people on site know that has not reached the project review?
The visit is useful only if the answers return to the cost forecast. A conversation without follow-up is pleasant. A conversation that changes the remaining-cost estimate, clarifies a contractual issue or triggers a decision is controlling. In my experience, the difficult part is often not discovering the problem. It is ensuring that the problem does not disappear between two departments.
There is a practical discipline to these conversations. Ask for the work still to be done, not merely the percentage complete. Ask what will prevent the next milestone from being achieved. Ask which activities have been performed but not yet captured in a time sheet, purchase order or invoice. Then compare the answers with the quantities and assumptions used in the estimate. The important word is ‘compare’. Without a link to the original estimating logic, even a detailed site visit produces observations rather than commercial control.
Two hours before the contract can save a great deal of explanation later
There is another important intervention, but it takes place at a different point in the project life cycle. Before an offer becomes a commitment, I favour a structured cross-functional review of the quantity framework: the Mengengerüst-Durchsprache, or MGD. Sales, engineering, manufacturing, assembly, purchasing and estimating examine the assumptions together. Not a ceremonial sign-off. A proper discussion. The participants should arrive with the quantity structure, the technical scope, installation assumptions and known uncertainties in front of them. Sales explains what has actually been promised. Engineering tests the technical quantities and interfaces. Manufacturing and assembly challenge the assumed effort and sequence. Purchasing tests material availability and supplier scope. Estimating makes the economic effect visible. The purpose is not to let six departments approve a number that nobody wishes to challenge; it is to make disagreements visible while the offer can still be corrected.
Where did the quantity come from? Who has allowed for interfaces? What is included in the installation scope? Is a performance requirement being interpreted in the same way by the sales team and the engineers? Have logistics, testing and commissioning been considered? Two hours can be enough to expose assumptions that would otherwise survive untouched until execution. A useful discussion moves from the largest and least secure assumptions to the interfaces that routinely disappear between disciplines. For example, a line item may include the machine itself but not the access equipment required for installation. A technically plausible assembly duration may depend on uninterrupted site access that the contract does not guarantee. A supplier quotation may cover delivery but exclude tests or documentation. The team should leave with corrected quantities, explicit assumptions, assigned open points and a traceable basis for the estimate. Otherwise the meeting has merely produced another set of minutes.
Over many years, an internal cost-estimating record — the MAK list — showed a reduction in average deviation between preliminary and final estimates from roughly 20 per cent to below 3 per cent in the industrial setting in which I worked. This is a documented practical observation, not a controlled scientific study and not a promise that every organisation will achieve the same result. Nevertheless, it illustrates what can happen when people challenge the quantities before they become contractual facts. The important lesson from that experience is the method, not a promise that every company will reproduce the same percentages. The review created a setting in which operational knowledge could change the estimate before a selling price and delivery obligation became fixed. It also made later differences easier to investigate: the team could distinguish a genuinely new event from an assumption that had never been tested. In my experience, that transparency is at least as valuable as the numerical improvement itself.
MGD and Walking Around are not duplicates. One tests assumptions before the contract; the other tests them against reality during execution. The first helps prevent avoidable surprises. The second helps recognise surprises early enough to act.
The meeting itself is not magic. Its value depends on preparation and the willingness of specialists to challenge one another. A quantity that engineering considers self-evident may include an installation assumption nobody has priced. Purchasing may have a quotation with exclusions that do not appear in the summary. Sales may have promised a delivery boundary that production interprets differently. A two-hour discussion can expose these differences precisely because the participants are looking at the same scope, not at six separate departmental presentations. The aim is not consensus at any price. It is to make assumptions explicit before they become contractual obligations.

Figure 2. Quantity review before contract and Controlling by Walking Around during execution.
Review explains yesterday. Preview changes tomorrow.
The standard project review begins with last month’s deviation and often ends with a lengthy explanation of why it occurred. Such explanations are necessary. They are not sufficient. A useful preview asks what is likely to happen between today and completion, what it means for the remaining costs, and which decision could still change the outcome.
Consider a fixed-price installation sold with an apparent limit of 60 days. What happens when the installation takes longer? Is the commercial scope genuinely capped? Is there a contractual basis for additional payment? Has the project forecast recognised the extra labour that may be required, irrespective of the outcome of the claim? A technically precise schedule is of little comfort if the contractual and economic assumptions contradict each other.
The management meeting should focus on a handful of issues that matter: cost to complete, open contractual exposures, credible recovery opportunities, the decisions required and the people responsible. Twenty slides of traffic lights are no substitute for one uncomfortable question asked at the right time.
An event-based preview also avoids waiting for the calendar. If a design change materially affects installation hours on Tuesday, the forecast should not have to wait for the last working day of the month. The accounting close has its rhythm; project decisions have another.
For a preview to be useful, every significant change should leave a short decision trail. What event triggered the revision? Which estimate assumption has changed? What is the new expected cost to complete? What can still be influenced, by whom and by when? At the next meeting, the first question should be whether the agreed action worked. This is where many organisations lose momentum: they document a risk, discuss it thoroughly and then treat the minutes as the result. Minutes do not reduce installation hours. Decisions, followed by action, sometimes do.

Figure 3. Event-driven preview: convert new information into decisions and an updated forecast.
When every department meets its target
A project can lose money even while the purchasing department reports savings, manufacturing meets its hour budget and engineering closes its tasks on schedule. The reason is simple: local success does not necessarily add up to project success.
A cheaper component can increase assembly time. A design released early can create expensive rework. A delivery recorded as complete can leave the installation team waiting for one essential item. These effects move across organisational boundaries. The cost does not vanish; it changes address.
Modern project cost control must therefore follow economic consequences across departments rather than merely confirm that each department has met its own target. The project, not the cost centre, is the unit whose result ultimately matters.
This also changes the meaning of responsibility. The objective is not to identify a culprit for every forecast change. It is to ensure that someone raises the issue, someone evaluates the consequence and someone makes a decision. If people learn that bad news is punished, management will eventually receive nothing but good news. The project result is unlikely to cooperate.
What software and AI can — and cannot — do
ERP systems are essential. They provide structure, traceability and reliable accounting records. But a system can only report the events it has been told about. If a bill of materials is closed while outstanding work remains unreported, the system may produce a perfectly consistent picture of an incomplete reality.
AI can help flag unusual cost patterns, conflicting completion notices, overdue purchase orders and schedule changes that deserve attention. It can support the comparison of quantities and the identification of potential gaps. Those are valuable uses. But an algorithm cannot settle a contractual ambiguity merely because it has found a suspicious pattern. Nor can it verify whether a crew still has three days of work on site without credible operational input.
The most important interface is still the one between people who understand the work and people who understand its economic consequences. Better software can strengthen that interface. It cannot replace it.
A sensible use of AI would be to bring potential inconsistencies to the surface: a work package marked complete while relevant purchase orders remain open; installation hours rising after mechanical completion; a forecast that has not changed despite a delayed milestone. These are prompts for investigation, not automatic proof of an error. The system does not know whether an open order is still required, whether a delay is recoverable or whether a reported completion refers to mechanical work or contractual acceptance. Those distinctions require the people responsible for the project. Technology can shorten the distance between a warning signal and a question. It cannot take responsibility for the answer.
The question I would add to every completion notice
Before accepting a completed work package as commercially closed, ask one final question: Is the work economically finished, or merely finished in the system?
If the answer is uncertain, record what remains, estimate its effect, assign responsibility and agree when the matter will be resolved. This is not another bureaucracy for its own sake. It is a way of preventing known costs from becoming late surprises.
Modern project cost control is not the art of explaining deviations after the fact. It is the discipline of recognising economic consequences while there is still time to do something about them. The numbers may be correct. The project forecast must also be complete.
Author note: Carsten Knappe-Finger is a graduate in business administration with more than 30 years of industrial experience in estimating, project management, engineering and productivity management. His work focuses on intelligent cost management and the practical connection between estimating, communication and project control.
Editorial disclosure: The MAK-list figures describe the author’s operational experience and should not be interpreted as a general statistical benchmark. No invented financial case study is used.
Written by Carsten Knappe-Finger.
Have you read?
How to Stop Bringing Workplace Stress Home.
Seven Healthy Aging Habits for Busy Women.
Why Singapore and Ireland Are Reshaping Global Wealth Geography.
How NASCAR’s Richest Drivers Built Fortunes Beyond Racing.
Lula Wins Fourth Term as Brazil Delivers Divided Mandate.
Follow CEOWORLD Magazine on: Google News, LinkedIn, X (Twitter), and Facebook.
Note: The views expressed are those of the authors and do not necessarily reflect those of CEOWORLD Magazine, its Editorial Board, or management. Content is provided "as is," may contain monetized links, and is not professional advice. See our Ethics & Guidelines for details. Reproduction requires prior written permission.
Contact info@ceoworld.biz for inquiries or media requests.