A stage gate that cannot return a no is a reporting event

Almost every large organization runs its capital projects through stage gates. Very few of those gates ever stop a project.

The pass rate is worth examining, because it has only two explanations. Either the development process is so reliable that no project ever reaches a gate underprepared, or the gate cannot return a negative answer. The second explanation fits the observed evidence better, and it is a governance failure that looks exactly like a functioning process.

What a gate is for

A gate exists to test readiness before further money is committed. Its function is to ask whether the work of the preceding phase has actually been completed to a standard that makes the next phase sensible — not whether the project is still wanted.

That makes it a quality control on information, not a referendum on the investment. A gate that asks "should we do this project" will always be answered yes by a room that has already spent money developing it. A gate that asks "is the design brief closed, is the utility position confirmed, is the estimate based on defined scope" can be answered no without anybody's commitment to the project being questioned.

The distinction is what makes a no survivable, and a no that is not survivable will not be given.

Why gates stop functioning

Four mechanisms account for most of it.

The date has already been committed. Where a completion date has been announced to a market, a regulator, a tenant, or a board, the gate is being asked to approve the only path that reaches it. Holding the project costs the date, so the gate approves and records concerns as actions.

Actions replace conditions. The most common failure mode is the pass-with-actions. The project proceeds, the actions are assigned, and nothing verifies closure before the next commitment. A condition that does not gate anything is a note.

The reviewers are the sponsors. Where the people at the gate are the same people accountable for delivering the project, the gate is a self-assessment. The information presented, the standard applied, and the decision taken all come from one interest.

Nobody has ever seen a hold. Organizational behavior is learned from precedent. In a system where no project has been held, presenting a project as not ready is an unusual act, and unusual acts are career-relevant.

A gate with a hundred percent pass rate is not a control. It is a reporting checkpoint that has been given the name of one.

What a working gate looks like

Gates that function share a small number of features, and none of them are elaborate.

The entry criteria are defined in advance and are objective. A named list of deliverables, each with a completion standard, produced before the phase began rather than assembled to match what exists.

The reviewers are independent of delivery. They may be internal, but they cannot be the project's own leadership, and at least one of them should have no stake in the project proceeding.

The outcomes include a real hold. Pass, hold pending defined evidence, or stop. Where hold is used, the project does not draw further funding until the evidence is produced — which is the only mechanism that makes the condition mean anything.

The decision is recorded with its basis. What was presented, what was tested, what was accepted on assumption, and what the reviewers relied on. A gate record that says only "approved" cannot be audited later, and later is when it matters.

The uncomfortable part

An organization that installs working gates should expect its first few to be unpleasant. Projects that would previously have passed will be held, dates will move, and the process will be described as bureaucratic by people whose schedules it has disrupted.

That is the cost of the control, and it is paid early and visibly, which is why it is unpopular. The alternative cost is paid late and is much larger: a project that proceeded through five gates on incomplete information, and arrives at construction with a design that assumed things nobody ever confirmed.

The test of a gate regime is not whether it is liked. It is whether anyone can name a project it stopped.