Success is currently declared at handover, against a forecast, by the people who made the forecast. The date, the criterion and the judge are all choices — and all three can be changed for nothing.
There is a footnote in a 1998 research paper that repays reading twice. The paper — Winch, Usmani and Edkins in Construction Management and Economics — spends fifteen pages arguing that the conventional triangle of time, cost and quality hampers understanding of what a project is for, and that success is better measured as the minimisation of client surprise. Then, to establish that the project it studies was in fact a success, it records that the client officially declared it so: completed on time, under budget, and to user satisfaction.
The authors are not careless. They have run into the thing every organisation runs into: when the moment comes to say whether something worked, the only evidence available on the day is conformance to a forecast, and the other kind — the kind their own argument says matters more — will not exist for years.
That problem is not philosophical. Most enterprises evaluate their largest investments at the one moment when success is least knowable, using a criterion that measures adherence to a prediction, judged by the people who made the prediction. Each of those three is a choice. Each can be made differently, at approval, at no cost.
The Strategic Context
Two sources thirty years old and on different continents arrive at the same conclusion by different routes, which is unusual enough to be worth noticing.
Derek Lidow, writing in the Project Management Journal in December 1999 as the chief executive of a manufacturer, proposed an alternative definition outright: successful change satisfies both its sponsors and its constituents, and is sustained by its constituents over time. He was explicit about the cost: it permits nobody to claim success during implementation, or even immediately on completion, and it displaces what he calls the holy grail of conventional practice — executional excellence, the idea that doing the thing well is the achievement.
Winch, Usmani and Edkins arrive at a compatible position from research. Their measure is the project performance gap: the distance between what the client expected at the outset and what they perceive they received. Their thesis is that a surprised client is a dissatisfied client. And they report two findings from other researchers that sharpen the point considerably. In a study of 646 projects, Baker and colleagues found that perceptions of the project outcome mattered more to success than meeting any particular objective. In a survey of 138 clients, Bresnen and Haslam found no direct correlation between client satisfaction and actual time and cost outcomes — but that performance relative to anticipated duration and cost was the baseline used for appraisal.
Read together, the finding is uncomfortable: satisfaction tracked the gap, not the outcome. A project landing where people were told it would land satisfies them; one landing in the same place having promised better does not. Both are secondary citations, reported inside the paper rather than read here, and both are indicative rather than definitive — but they point the same way as Lidow's definition, and neither source knew of the other.
What Leaders Commonly Misread
That the criterion is fixed by nature. On time, on budget, to specification feels less like a choice than like the definition of the word. It is a choice, and a fairly recent one, and it has a specific property that makes it attractive independent of its truth: it is the only criterion that can be evaluated on the day everyone is still in the room.
That declaring success is a reporting act. It is a governance act with consequences. It determines which initiatives are held up as models, which sponsors are promoted, which delivery approaches are repeated, and what the organisation believes about itself. An enterprise that has declared success on twenty initiatives whose value never materialised has not merely misreported; it has trained itself.
That the people who ran it are the people to ask. Lidow's formulation puts the verdict partly in the hands of constituents — the people who must live with the change and whose continued use of it is the evidence that it worked. That is a different population from the sponsor, and usually a different one from the users consulted during design. They are rarely asked, and when they are asked it is through a satisfaction survey issued in the fortnight after go-live, which measures relief rather than value.
That a good process guarantees a good verdict. Lidow argues that conventional practice leaves the impression that mistakes have been made whenever a schedule slips or a budget is exceeded by so much as a penny — and holds instead that any change should be considered well executed if it turns out to have been successful. That inverts the usual logic, judging execution by outcome rather than outcome by execution. It is an uncomfortable inversion for an assurance function, and closer to how boards actually reason after the fact.
Reframing the Issue
Treat the verdict itself as a designed instrument with three settings, all fixed at approval and all currently set by default.
The criterion. What condition, observable in the world, will count as this having worked? Not a milestone; a state. Trains running with the reliability the case assumed. A collection actually being used by researchers. Residents staying longer, or staff staying at all.
The judge. Who is entitled to declare that condition met? The sponsor is the obvious candidate and the worst one, because the sponsor is the author of the forecast being tested. The candidate that follows from both sources is the constituency — the operating function, the users, the people whose continued behaviour is the evidence.
The date. When will that party be asked? A date far enough out that the condition could plausibly have appeared, and near enough that someone still remembers what was promised. It should be in the approval paper, in the diary, and funded.
Note what this is not. It is not the design of leading indicators, nor how quickly a deviation can be detected during delivery — that is [Related article: Measuring an Outcome You Cannot Predict], and no claim about detection is made here. Nor is it who sits on the governing body while work is under way, which [Related article: Who Sits on the Board for the Benefits?] addresses. This is the verdict after that body has dissolved.
Strategic Analysis
The evaluation window closes exactly where the money starts
There is a structural reason the current arrangement persists, and it is written into how the work is scoped in the first place.
Course material on project cost management states it plainly: non-acquisition costs usually fall outside the scope of the defined project. The boundary is drawn at handover — which is to say, drawn immediately before the period in which the great majority of an asset's cost is incurred. The same material acknowledges the trade-off this creates, noting that extra design effort may generate operations and maintenance savings, and that a project manager should be willing to adjust the work to realise them when justified. Winch and colleagues make the same point from the quality side, naming life-cycle costing as an instrument for managing the fitness of a specification rather than its execution.
The willingness is not the difficulty. The measurement is. A delivery leader asked to spend more capital now for operating savings later is being asked to worsen the only number they will be judged on, in exchange for a benefit that appears after their evaluation has closed, on someone else's budget line. That is not a character problem. It is an evaluation-window problem, and it is fixed by moving the window rather than by exhorting people.
A rolling-stock replacement makes this concrete, hypothetically. Two bids differ by nine per cent on capital; the dearer one has better maintenance access and a longer overhaul interval. Every instrument in the approval pack favours the cheaper bid. The only one that would favour the other is a whole-life comparison, and the only reason to build one is that somebody will still be asking in year seven. Set the verdict date there and the case gets built. Leave it at handover and it does not, whatever the policy says.
What the enterprise currently does at the end, and why it does not count
Most organisations do run a closing ritual, and the teaching material describes it accurately. At finalisation there is a final cost variance report against the budget at completion, to be widely disseminated. On the quality side there is a summary of quality results, a final customer satisfaction review, and lessons fed into the corporate knowledge base.
Every element is conformance to a forecast, or an impression collected while the delivery team is still present. None can answer whether the change was sustained, because nothing has had time to be sustained or abandoned. It is a competent close-out, not a verdict — and treating it as one is how organisations come to believe things about themselves that are not true.
The one thing an enterprise gets in exchange
There is a real objection to all of this, and it should be stated rather than dodged: a verdict at year three is useless for managing the work at month six. Delivery needs feedback on a delivery timescale, and no organisation can run on retrospective judgement alone.
That is correct, and it is why this is an addition rather than a substitution. Keep the close-out; keep the in-flight measurement. Add one instrument that closes a loop nothing else closes — between what an investment promised and what the organisation actually got, judged by people with no stake in the answer.
Its value is not accountability, and framing it as accountability will kill it. Its value is calibration. An enterprise that has run the verdict twenty times knows which of its cases were optimistic, which sponsors forecast well, and which classes of investment reliably underdeliver. None of that is knowable from a close-out report, and all of it changes capital allocation.
Decision Framework
The instrument is a verdict clause: five lines in the approval paper, agreed before funding.
| Line | The commitment |
|---|---|
| The condition | The observable state that will count as success — expressed as something happening in the world, not as a deliverable existing |
| The judge | The named party entitled to declare it, chosen from those who must live with the result rather than those who built it |
| The date | When they will be asked. Far enough for the condition to appear; near enough that the promise is remembered |
| The funding | What the evaluation itself will cost, and from which budget. An unfunded evaluation does not happen |
| The consequence | What the organisation will do with an unfavourable verdict — which must be something other than nothing |
Three tests.
The absent-sponsor test. Would the verdict be the same if the sponsor were not in the room? If the answer depends on who is present, the judge has not been chosen; it has been defaulted to whoever convenes the meeting.
The year-three test. Could the condition conceivably be false three years after handover while every delivery metric was green? If not, it is a restatement of the delivery plan and no new information will ever come from it.
The stated-versus-assumed test. Dissatisfaction often arises not between promise and delivery but between what was written and what was understood. [Related article: Silence Reads as a Promise] deals with expectations created by what a scope statement omits; the point here is narrower — before setting the condition, ask the receiving function what they believe this will do for them, and reconcile the two answers in writing while that is still free.
From Strategy to Execution
Immediately. Add the verdict clause to the next three investment papers. It is five lines, and the argument it starts is worth more than the clause.
Over two or three quarters. Fund the evaluations and put them in a diary that outlives the programme. This is the step that fails: the clause is written, the sponsor moves on, and nobody owns the appointment. Ownership belongs with whoever governs the portfolio, not with the initiative, because the initiative will not exist.
Longer term. Build the record. Twenty verdicts, held centrally and compared against their original cases, is a calibration asset no consultancy can supply and no competitor can copy. A separate question — who owns the benefit once a temporary structure dissolves — belongs to [Related article: The Hidden Cost of Putting Work Into Project Form].
Signals to Monitor
- Success declared in the same month as handover. Definitionally, the evidence cannot yet exist.
- Post-implementation reviews scheduled but not funded. The reliable predictor that they will not occur.
- A track record of successes alongside flat enterprise performance. The most important signal in the list, and the hardest to raise.
Questions for the Leadership Team
- Who declared our last three major investments successful, on what date, and against what condition?
- If we asked the operating function that inherited each of them today, what would they say?
- What observable state would tell us, three years from now, that our current flagship investment worked?
- Who is entitled to give that verdict, and is it funded and diarised?
- What would we do with an unfavourable verdict — and has that ever happened here?
Closing Perspective
An organisation cannot learn from a judgement made too early, by the wrong people, against the wrong thing. It accumulates a record of well-run projects while wondering why the enterprise is not obviously better off.
The correction is cheap: no new function, no system, no methodology — five lines in an approval paper, a diarised appointment, and the willingness to let someone other than the author mark the work. What makes it hard is not cost but exposure. An enterprise adopting the verdict clause is choosing to find out, on the record, how many of its confident cases were right.
It is the same choice in four other forms: a specification nothing can question — [Related article: Quality Assurance Cannot Tell You the Specification Was Wrong]; a priority label nobody prices — [Related article: What Does "Critical" Actually Commit You To?]; a number stripped of its range — [Related article: From Estimate to Commitment]; a process nobody can see into — [Related article: Visibility You Cannot Use]. In each, the instrument answers a narrower question than the enterprise believes it is asking, and the remedy begins with saying so out loud.
About EraNorth Insights
EraNorth Insights publishes practical analysis on strategy, projects, operations, transformation and decision intelligence for professional and organisational use. About EraNorth.