Nobody sets out to build a bureaucracy. Every heavy stage-gate process started as three good intentions: get bad projects killed early, get good projects through fast, and keep a clean record of why each call was made. Then it grew a review board nobody remembers approving, and the good projects started waiting as long as the bad ones.
That is the actual failure. Not that gates exist. That almost nobody still measures them against the job they were built to do.
The Test a Stage Gate Was Actually Built to Pass
Governance run properly delivers three things, and nothing else matters as much as these: faster decisions, so a good project stops waiting weeks for a yes. Earlier kills, so a weak one frees up capital instead of quietly draining it for another two quarters. And a clean audit trail, so nobody has to reconstruct the reasoning after the fact. Get those three right and a stage gate speeds decisions. Miss them and it slows every decision down, good and bad alike, because the mechanism has stopped doing the job it was built for.
Someone genuinely has to decide which projects live, which pivot, and which ones are quietly draining the business. That much was never in question.
Why the Mechanism Rots
Four patterns do most of the damage, and each one accumulates quietly rather than arriving as a single bad decision. Entry gates get heavy while exit and kill discipline stay weak or disappear entirely, so zombie projects clog the funnel and starve the strong ones of attention. Panels grow larger and meetings grow longer until authority is spread across so many people that nothing actually gets decided. Reviews turn into theatre, rubber-stamping or deferring rather than choosing, and momentum dies in the gap between gates. And decision rights stay ambiguous enough that nobody is the clear owner of the yes or the no, so everything escalates and stalls at once.
Each of those four is a governance design that stopped serving the teams running through it and started serving itself, not a process flaw you fix by adding another step. The entry gate feels productive, so it keeps growing. The exit gate feels harsh, so nobody wants to own it, and that imbalance is where most of the trouble actually starts.
The Numbers Behind the Frustration
The frustration shows up in real operating numbers, not just complaints. At Vivix Vidros Planos, a Brazilian flat-glass manufacturer, resolving a customer complaint used to take weeks, long enough for a buyer to lose patience and take the next contract elsewhere. After building an AI-powered chatbot into its existing production and quality data, that shrank to minutes, an 80% reduction in complaint resolution time. Responses to production-line issues sped up by a further 85%. Both figures come from Vivix’s own case study, published jointly by Siemens and AWS. The underlying pattern holds regardless of the exact numbers: the real cost of a slow gate shows up as lost trust and lost contracts, not just lost hours.
What Minimum Viable Governance Actually Looks Like
The fix is sizing each gate to the actual risk in front of it, rather than running every decision through the same heavy process regardless of what it actually requires. A low-risk process tweak does not need the same panel as a bet-the-quarter platform launch, and treating both the same is exactly how standing committees fill up with work they should never see in the first place. High-risk decisions keep full board review, because that rigor is proportionate there. Mid-tier decisions get a lightweight gate with a single accountable owner. Low-risk work proceeds by default through a simple intake form, with oversight applied only if something in it actually warrants it.
The useful design target sits between two failure modes: above a ceiling, controls become bottlenecks and teams quietly route around them; below a floor, real risk creeps in unmanaged. Getting that band right is not abstract. One organisation that tightened its policy down to the minimum viable version halved the time complex decisions took and surfaced three times more opportunities than peers still running the heavier version.
The Test Most Gates Would Fail
Pick the last three projects your organisation killed at a gate, and the last three it approved. If the kills took longer to reach than the approvals, the gate is not protecting the business from bad decisions. It is protecting itself from having to make any decision at all.