A gate that has never been failed is not a control

There is one question that establishes whether a stage gate is a control or a ceremony, and it takes about a minute to ask.

How many have been failed?

Not deferred. Not passed with conditions, unless somebody can also show that the conditions were closed. Failed, in the sense that the programme was told it could not proceed and did not proceed.

On a great many capital programmes the honest answer is none, ever, and the people answering do not experience this as a problem. The gate happened. Papers were issued, a committee sat, minutes were taken. What did not happen was the possibility of a negative answer, and a control that cannot return a negative answer is a reporting event wearing the clothes of a control.

Where the leverage actually is

The case for examining decisions early is not a moral one about rigour. It is arithmetic.

The cost of changing a decision rises through the lifecycle, steeply and without exception. A specification changed during concept costs the time of the people in the room. The same change during detailed design costs rework across every discipline that has already consumed it. During construction it costs a variation, and after handover it costs an alteration to something that is already in use, plus the cost of not having had it for however long.

FIG. 01The relationship holds regardless of the absolute numbers. What matters for assurance is the shape: the cheapest place to change a decision is the place where the least evidence exists to justify changing it.

So the stage where assurance can change the most is the stage where it is cheapest to act on, and that is where almost none of it is spent.

Why the money goes elsewhere, and it is not carelessness

The usual explanation is that owners underinvest in early work. That is true and it is not the mechanism.

Assurance functions examine artefacts. That is what they are built to do: read the cost report, test the schedule logic, sample the change register, trace a payment to its authorisation. Every one of those requires the artefact to exist.

Inception produces almost none of them. There is no cost report, because there is no cost yet. There is no programme in any meaningful sense. There is no site, no contractor, no earned value, no register of anything. What exists is a set of assumptions held by a small number of people, some of it written down and much of it not.

An assurance function arrives when the artefacts arrive, which is always after the decisions those artefacts record.

This is why early assurance feels unbuyable rather than merely unbought. The standard offering does not fit the stage. Asked to review a programme at concept, a conventional assurance team will find very little it recognises, and will produce a report about governance arrangements rather than about the decision, because governance arrangements are the only artefact available.

What is actually examinable before anything is built

Quite a lot, but it requires asking different questions rather than asking the usual questions earlier.

Whether an outcome exists separately from a solution. Most sanction papers describe a thing to be built. Rather fewer describe, separately and measurably, what that thing is for after it is finished. Where no such statement survives, every later question collapses into whether the thing was built as drawn, and it becomes impossible to ask whether it should have been that thing.

Whether the criteria were written before the appraisal. An option appraisal without criteria recorded in advance cannot be audited. There is no standard against which the choice could be wrong, so a reviewer can only substitute their preference for the sponsor's. This has recently stopped being a maturity observation and become a nonconformity, because the 2024 edition of ISO 55001 requires the criteria to be documented.

Whether the estimate declares its class, and whether the class is true. An estimate class is a statement about how much of the scope was defined, not about how hard anyone worked. A number produced at low definition and carried into a sanction paper without its range commits the organisation to a figure that was never a commitment.

Whether anybody has priced the ownership. Capital and operating budgets usually sit in different organisations, and every design decision that fixes operating cost is taken inside the first. If no design decision on the programme has ever been altered on operating cost grounds, whole-life costing is presentational.

Whether the information requirement was set before the design that produces the information. Requirements issued afterwards cannot change what is captured, only what is compiled.

Each of those has a checkable answer, and none requires an artefact that does not exist yet. The framework for this stage sets out all six domains and the published clause behind each.

Three conditions that make a gate real

If the gate is the mechanism, and on most programmes it is, then it is worth being precise about what separates one that works from one that does not.

The criteria were set before the submission was written. Where the submission comes first and the criteria are inferred from it, the gate is a review of internal consistency.

The chair does not carry delivery accountability. Where the person chairing the gate is also answerable for the schedule the gate might interrupt, the control and the thing being controlled share an owner. This is not an allegation about anybody's integrity. It is a structural observation, and it is visible on an organisation chart.

The committee sees a range rather than a point. A single number presented to a board becomes the number the board remembers, and every subsequent movement is reported as a variance against a figure that was never a forecast.

Why this matters more here than the general case

The Gulf has spent a decade sanctioning capital programmes at a scale and pace with few historical parallels, and much of that sanctioning happened under conditions that compress exactly the work described above.

Programme announcements frequently preceded scope definition rather than following it, which inverts the normal relationship between an estimate and a commitment. Where a figure is public before the definition supporting it exists, the estimate is no longer an input to a decision. It is a constraint the definition has to be made to fit.

The Public Investment Fund approved a minimum twenty per cent spending reduction across its portfolio in December 2024, and several giga-project timelines have been recalibrated since. That environment does not create the inception problem, it reveals it. A portfolio being reduced has to choose what to stop, and choosing requires precisely the criteria that were not written down at the point each programme was sanctioned.

What the estimate was actually telling you

The organisations that will find this easiest are the ones that can produce, for each programme, a statement of the outcome, the criteria under which the option was chosen, and the definition maturity the estimate rested on. That is a small amount of paper. It is not retrievable afterwards if it was never written.

Sources. AACE International Recommended Practice 56R-08, cost estimate classification for the building and general construction industries, revision of 7 August 2020, and 18R-97 for the process industries. ISO 55001:2024, Clause 4.5, asset management decision-making, new in the second edition, ISO standard 83054. ISO 21502:2020, guidance on project management, Clause 6 on the project governance framework. PMBOK Guide, Eighth Edition, November 2025, which carries governance and finance among its seven performance domains. ISO 19650-3, information management during the operational phase. The Public Investment Fund spending reduction was reported in December 2024. Standards published behind a paywall are cited without a link.