The Gate That Stopped Asking Questions
The gate was always meant to be a small quantity of hard thinking at a well-chosen moment — not a large quantity of documentation at a scheduled one.
The Thursday Everyone Already Knows
The gate review sits in the calendar for Thursday. By Tuesday, everyone who will be in the room already knows how it ends. The project will pass. The pack has been assembled — a hundred-odd pages of plan, updated risk log, a refreshed business case, and one slide showing the milestones nudged three weeks to the right with a footnote explaining the slippage. The reviewers will ask their questions, the project manager will give the answers rehearsed at Monday’s dry run, and the status will be marked green. The only real uncertainty in the week is whether the meeting starts on time.
I have sat on both sides of that table often enough to know the ritual for what it has become. Somewhere in the last few years, across a great many organisations, the stage gate stopped being the moment we decided whether to go on and became the moment we confirmed that we already had. It still costs what a decision costs — the elapsed weeks, the assembled pack, the diaries of senior people — but it no longer returns what a decision returns. It adds delay without adding insight. That is a poor trade, and we have been making it quietly and at scale, while congratulating ourselves on our discipline.
What a Gate Is Actually For
A stage gate is not a document check. It was never meant to be. Its purpose is to force a single, uncomfortable question at a defined point: on what we now know, is this still worth doing, on this basis, at this cost? The gate exists because that question is easy to avoid. Once a programme is moving, momentum, sunk cost, and the reputations riding on it all conspire to keep it moving. The gate is a deliberate interruption — a place where the organisation gives itself permission to stop, or to change course, before the next tranche of money is committed.
The discipline came to most of us from good places. The tollgate logic of the product-development world, the Gateway reviews that central government now runs on its major programmes, the stage boundaries written into the method manuals on every project manager’s shelf — all of them encode the same sound instinct, that large commitments should be released in stages against evidence rather than in one act of faith at the outset. The instinct was right. What has happened to it in practice is the subject of this piece.
The mechanism only works if the answer can genuinely be “no”, or “not like this”. A gate that can only return “yes” is not a gate. It is a milestone wearing the costume of a control. And the tell is simple: if nobody in the room can remember the last time a review changed a decision — sent a project back, reshaped its scope, paused it for a hard look — then the gate has stopped doing the one thing it is for.
A gate that can only return “yes” is not a gate. It is a milestone wearing the costume of a control.
How the Question Drained Out
No one decides to hollow out a gate. It happens by increments, each of them reasonable on the day it is made.
- The pack grows. A review once asked for a business case and a plan. Then a governance framework was adopted, and the pack acquired a template. Then an incident on another programme prompted a new mandatory section, and the section stayed. Each addition was defensible; the sum is a document so large that assembling it consumes the very analysis it was meant to contain. The team spends the fortnight before the gate producing the pack, not interrogating the project.
- The reviewer loses the authority to stop. In the original design, the gate is chaired by someone with the standing to halt the programme. In practice, seniority drifts downward, or the chair turns out to be the same executive who sponsored the project and cannot easily gainsay their own business case. A reviewer who cannot say “no” can only witness.
- The calendar takes over. Because the gate is a fixed date, it becomes a scheduling event. The live question stops being “should this proceed?” and becomes “will the pack be ready by Thursday?” Elapsed time is spent, but on logistics rather than judgement.
- The green becomes self-protective. Everyone in the chain has an interest in the gate passing. A red status invites scrutiny of the reviewer as much as the reviewed. The path of least resistance is amber-trending-green with a remediation note — and the path of least resistance is the one a tired system takes every time.
Consider a composite that will be familiar. A large back-office replacement reaches its second major gate. The pack runs to a hundred and twenty pages and takes three people the best part of three weeks to compile. Once the diary-finding and the reworking of slides are counted, the review adds six weeks of elapsed time to the schedule. It passes, green, with four “minor” actions logged. What the pack never surfaces — because no section asked for it — is that the transaction-volume assumption underpinning the entire benefits case was set eighteen months earlier and has since been overtaken by events; the business the case was written for has quietly changed shape. The gate examined whether the documents were present and internally consistent. It did not ask whether the thing was still true. Two gates later, when the assumption finally breaks the surface, the cost of unwinding is an order of magnitude larger than a single hard question would have cost in that room.
That is the precise failure. Not that the gate found nothing wrong, but that it was no longer built to look.
Two Kinds of Green
It helps to separate the two things a green gate can mean, because we have let them blur into one colour. There is informative green: we asked the difficult question, tested the assumption that could kill the programme, and satisfied ourselves that it holds — proceed. And there is performative green: the pack was complete, the template fields were filled, the meeting happened on schedule — proceed. On the status report they look identical. In their value to the organisation they are worlds apart. Informative green is a decision. Performative green is an administrative event that has learned to impersonate one. When a portfolio’s gates are almost all green, the honest question is not “why are our projects so healthy?” but “which kind of green are we actually producing?”
The Case for the Gate — and Where It Holds
It would be easy to read all this as an argument against gates, and to reach for the fashionable conclusion that governance is overhead and delivery should simply be set free. That is the wrong lesson, and the strongest version of the opposing case deserves a fair hearing rather than a caricature.
Gates do catch bad projects. For every review that rubber-stamps, there is one that — precisely because a date forced the organisation to look — surfaced a dependency no one owned, or a benefits case that evaporated under three good questions. Discipline is not the enemy of delivery; the absence of it is how organisations sleepwalk into commitments they cannot honour. Remove the gates entirely and you do not get faster, wiser programmes. You get programmes that never have to answer for themselves until the money is already spent. The instinct that built the gates was sound, and the teams straining under badly run reviews are often the last people who should be trusted to abolish them.
So the argument is not against the gate. It is against the ceremony that has replaced it — and the two have grown so alike on the surface that we have stopped telling them apart. The task is not demolition. It is restoration.
“The problem is not that we hold too many gates. It is that we hold too few real ones.”
Restoring the Question
The repair is less about redesigning the process than about returning to it the thing it has lost. A few disciplines, in my experience, do most of the work.
- Give every gate one live question. Before the review is booked, name the single decision it exists to inform and the one assumption that, if it has failed, should stop the programme. If no such question can be named, the gate is not needed and should be cancelled — which is itself a useful finding, not an embarrassment.
- Shrink the pack until it fits the question. A review pack should be the evidence for that decision and almost nothing else. If a section does not bear on whether to proceed, it belongs in the project’s records, not in the gate. A shorter pack is not a weaker control; it is a sharper one, because there is nowhere for the hard number to hide.
- Seat someone who can say no — and who does not already own the answer. The reviewer needs both the authority to halt and enough distance from the business case to want to test it. A degree of studied ignorance in the chair is an asset: it forces the obvious question that the insiders have long since stopped asking of themselves.
- Measure gates by decisions changed, not projects passed. A gate process with a hundred per cent pass rate is not evidence of excellent projects; it is evidence of a control that has stopped controlling. Track how often reviews actually alter a course. When that figure is zero across a whole portfolio, the ceremony has fully displaced the decision, and no amount of tidier documentation will bring it back.
None of this adds process. Most of it removes process and replaces it with attention. The gate was always meant to be a small quantity of hard thinking at a well-chosen moment — not a large quantity of documentation at a scheduled one. We have spent years perfecting the documentation. The thinking is what we owe the next programme.
So the next time a review is booked for a Thursday and the room already knows the answer by Tuesday, that foreknowledge is not a sign the project is healthy. It is the gate telling us it has nothing left to ask. The right response is not to run the ceremony more smoothly. It is to make it dangerous again.