Quality gates in production: where the flow stops and who is allowed to release it

A gate stops flow; an inspection only produces a result

Plants frequently believe they have gates when what they have is measurement. If a check records a value and the parts move on regardless while somebody reviews the paperwork later, no gate exists at that point. Making it a gate requires a defined halt, an authority to release, and a physical or system constraint that makes continuing without release difficult. That constraint is the part usually missing: without it, the gate depends on the goodwill of whoever is under pressure to hit the day's numbers, which is exactly the circumstance the gate was placed to survive.

Place them where value is about to be added irreversibly

The economics favour stopping immediately before an operation that would make the defect expensive or impossible to recover: before a costly machining stage, before assembly into a larger unit, before potting or welding that prevents disassembly, before painting or plating, and before dispatch. A gate placed after such a step still finds the defect, but by then the material and the labour are lost. Mapping where the accumulated cost jumps, and putting the stop just ahead of each jump, produces a far better layout of gates than distributing them evenly along the route.

Release authority has to be specific and unwelcome to bypass

Name the role that can release at each gate, and name it narrowly enough that everyone knows who it is on the night shift. Where a release requires a concession because the product does not fully conform, that authority should sit at a different and higher level than routine release, since the two decisions carry different consequences. Recording every release with the person, the evidence and the time is what makes the arrangement auditable and, more usefully, makes the pattern of who releases what visible when a defect eventually escapes.

Gates accumulate, and each one costs flow

Every gate adds queueing, handling, paperwork and a decision-maker who may not be immediately available, so a line with many of them will develop informal routes around the ones that hurt most. Gates also tend to be added after incidents and rarely removed, so their number grows year on year while the original reasons are forgotten. A periodic review should ask, for each gate, what it has actually caught over the past period, whether an upstream control now prevents that defect, and whether the check could move into the operation rather than sitting between operations.

What happens on a fail decides whether the gate works

A gate that stops product without a defined next step creates a pile of held material and pressure to release it. Each gate needs its reaction defined: where the product physically goes, who is notified, how far back the suspect boundary extends, whether the line continues or stops, and what evidence is needed to resume. Held stock with no owner and no deadline is the single most reliable predictor that a gate will eventually be bypassed, because the operational cost of the hold falls on people who did not make the decision.

Frequently asked questions

Should the operator or a separate inspector hold the release at a gate?

It depends on what the gate protects. For routine dimensional confirmation where the criterion is unambiguous and the measurement is objective, the operator releasing their own work is efficient and works well. For a judgement call, for a characteristic with safety consequences, or where the operator is under a production incentive that conflicts with holding, independence is worth its cost. Mixed schemes are normal, and the reasoning for each gate should be written rather than assumed.

How do we stop gates being bypassed when we are behind schedule?

Make bypassing require a decision by someone senior rather than an omission by someone junior, and make the record of it visible the next morning. Physical constraints help — material that cannot leave an area without a released label, a system that will not book the next operation until the previous release exists. Equally important is removing the incentive to bypass by making the gate fast: most bypasses happen because the gatekeeper is unavailable, not because anyone disagreed with the gate.

Is a final gate before dispatch enough on its own?

It protects the customer better than nothing and it protects your margin very poorly, because everything caught there has already consumed the full cost of manufacture. A single terminal gate also concentrates all detection into one check under the greatest time pressure of the whole process. Earlier gates cost less to operate and prevent value being added to product that will not ship, so the dispatch gate should be the last line rather than the only one.