When stage-gate stops killing things, it stops being stage-gate

Take any stage-gate process and look at what proportion of projects die at each gate. Where that number sits below one in five, the gate has stopped functioning as a gate and become a reporting ceremony that happens to involve senior people.

The failure has a recognisable shape. Forty-five projects arrive at the quarterly review. Forty-three are green or amber. Two are red, one of which was red last quarter as well, and the other is being re-baselined. Nothing is killed, nothing is meaningfully re-prioritised, and the meeting adjourns having documented that the pipeline is healthy. What it has actually documented is that the pipeline is full.

Then, roughly once a year and usually shortly after a board asks for more discipline, something does get killed. It tends to be a project whose sponsor has recently left the business, or one small enough that killing it costs nobody anything. The kill is real, the discipline is performed, and the projects that most needed stopping are untouched. The signature is easy to spot once you look for it, and it tells you the gate is optimising for the appearance of rigour rather than for capital allocation.

Why the criteria don’t help

Most gate criteria are written as thresholds to clear rather than as predictions to test, and that difference decides how the conversation goes.

A threshold invites negotiation. The team did not quite hit the customer validation target, but the sample was unrepresentative because of seasonality, and the trend is encouraging. Every word of that may be true. It is also unfalsifiable in the room, which means the gate has nothing to push against.

A prediction does not offer that escape. The team said customers would pay a given price for a given outcome. The experiment says they will not. The bet was wrong, which is different from the team being wrong, and it is a claim a gate can actually adjudicate. Framed that way, the question stops being whether to keep funding a team and becomes whether the original investment thesis survives. If it does not, the follow-on question is whether the new thesis deserves a fresh allocation on its own merits, which is a much harder thing to wave through.

Three changes that move the number

The first is a pre-mortem written at the start of each phase rather than after the failure. One page, signed by the project lead and the gate sponsor, naming the three things that would mean this project should not pass its next gate. At that gate, reading it aloud is the first agenda item. The effect is mechanical rather than motivational: it forces the meeting to discuss failure modes it has already agreed are plausible, at precisely the moment it would otherwise be moving on.

The second is inverting the default. Most gates are built so that a project continues unless someone actively kills it, which puts the burden of proof on the person willing to be unpopular. Reverse it, so the project stops at the gate unless the sponsor actively re-funds the next phase and says which criteria were met. Nothing about the criteria changes. Decisions are sticky in the direction of the default, and moving the default moves the outcome.

The third is publishing the kill rate per gate, by quarter, alongside the names of the sponsors. This one is uncomfortable and it works for an uncomfortable reason: nobody wants to be the sponsor whose pipeline is conspicuously the laxest in the portfolio. It is social pressure rather than governance, and it should be treated as a corrective rather than a permanent fixture.

None of this is free either. A gate that starts killing things generates arguments, and some of them will be with people whose projects deserved to survive. That is the cost of a process that discriminates at all. A gate that never rejects anything has simply moved the argument somewhere less visible, usually to whoever is trying to staff the next quarter with the same people already committed to forty-three live projects.

If you’re staring at a portfolio review next quarter and you suspect this is your story, the half-day diagnostic is something we run as a fixed-fee engagement. See the engagement formats or get in touch.