Quality Gates in Manufacturing

A quality gate is a fixed checkpoint between two stages of a process where work must meet defined criteria before it is allowed to advance. At each gate the work is reviewed against a checklist, and the outcome is a decision, go, go with conditions, hold, or stop, so a defect or an incomplete deliverable cannot move to the next stage and pick up cost.

The idea is simple and old: it is cheaper to catch a problem where it is made than three steps later. Quality gates turn that instinct into structure. Instead of hoping problems get noticed, you build hard checkpoints into the flow where nothing passes until it earns the right to. This guide covers what a quality gate is, what criteria live at one, how the gate decision and its escalation work, how to set gates up across a process, and how they differ from ordinary inspection.

What is a quality gate?

A quality gate, also called a phase gate or stage gate, is a control point at the boundary between two stages where progress is checked against predefined criteria before the next stage can start. The concept came out of manufacturing and product development as a way to manage risk in new-product programs: break the work into stages, and put a review between each one where a go/no-go decision is made (ASQ, Quality Glossary). The goal is to find and fix problems early, before the effort and cost of the next stage are committed.

Two things make a gate a gate rather than a status meeting. First, it has criteria decided in advance a specific list of what must be true to pass, not a general vibe check. Second, it has authority the gate can actually stop the work, not just note a concern and wave it through. A checkpoint with criteria but no power to stop is a report. A checkpoint with power but no criteria is a gut call. A quality gate needs both.

Gates apply at two scales. Across a development program, gates sit between phases, concept, design, validation, launch, and this is exactly what Advanced Product Quality Planning (APQP) formalizes with its phase reviews. On a production line, gates sit between operations, a part cannot move from machining to assembly until it passes the gate between them. Same logic, different granularity: nothing advances until it meets the bar.

What criteria live at a quality gate?

A quality gate runs on three things: entry criteria, the deliverables and evidence to review, and exit criteria. Entry criteria say what must be ready before the gate review even happens, you do not convene a gate to look at half-finished work. The deliverables are the concrete evidence: drawings, test results, a completed first article inspection an updated control plan. Exit criteria are the bar the work must clear to pass.

The criteria have to be specific and checkable, because the whole point is to remove judgment from the moment of decision. "Design is basically done" is not a gate criterion; "all critical characteristics have tolerances, the DFMEA is closed, and the prototype passed validation testing" is. The table shows what gate criteria look like across a typical product program.

How does the gate decision and escalation work?

At a quality gate the reviewers compare the work to the exit criteria and reach one of four outcomes, and each has a defined path. This is what keeps a gate from collapsing into "close enough, let it through."

• Go. All criteria are met. The work advances to the next stage. This is the clean pass.
• Go with conditions. The work advances, but with specific open items to close on a committed date. Conditional passes are useful but dangerous, they only work if the conditions are tracked and actually closed, not forgotten the moment the work moves on.
• Hold. Criteria are not met; the work stays in the current stage until they are. The gate does its core job here, stopping incomplete or defective work from advancing.
• Stop. The work has a problem serious enough that continuing is the wrong call, a failed validation, a safety issue, a business case that no longer holds. Killing work at a gate is a feature, not a failure; it is cheaper than launching something that should not exist.

Escalation is what makes the hard outcomes stick. A gate needs a defined authority level for each decision, and the harder the call, the higher it goes. A routine go can be signed at the gate; a stop or a costly conditional pass should require a defined approver, a quality manager, a program lead, or above, so no single person under schedule pressure can quietly wave a defect through. Without an escalation path, gates erode: the first time someone overrides a hold to hit a date and nothing happens, the gate stops meaning anything.

How much does a gate save by catching problems early?

The economic case for gates is the escalating cost of a defect the further it travels. The widely cited 1-10-100 rule holds that a defect prevented in design costs on the order of one unit, catching it in production costs roughly ten, and letting it reach the customer costs about a hundred. The aggregate shows up in the cost of quality: ASQ reports that the total cost of poor quality can reach 15% to 20% of sales revenue in organizations that manage quality reactively (ASQ, Cost of Quality). Structured phase reviews are the standard tool for catching problems at the low-cost end of that curve, which is why frameworks like APQP build formal gate reviews into every product program (AIAG, APQP & Control Plan). A gate is cheap. A recall is not.

How do you set up quality gates?

You set up quality gates by deciding where the boundaries are, what has to be true to cross each one, and who can say no, before the work starts. The sequence:

1. Split the process into stages with clear boundaries. Identify the natural break points where work changes hands or risk jumps, concept to design, design to validation, machining to assembly. A gate belongs at each boundary where letting a problem through would add real cost.
2. Define exit criteria for each gate. Write the specific, checkable list of what must be true to pass, deliverables complete, tests passed, reviews closed. If a criterion cannot be verified with evidence, rewrite it until it can.
3. Assign a gatekeeper and an escalation path. Name who runs each gate and who must approve the hard outcomes. The authority to hold or stop has to sit with someone who is not the same person under pressure to hit the date.
4. Build the gate into the schedule, not on top of it. Treat gate reviews as committed events with time allotted, so they are not skipped when the program runs late. A gate that gets dropped under schedule pressure was never a gate.
5. Capture the decision and the evidence. Record the outcome, the criteria met, any conditions, and the sign-off. This turns the gate into a traceable record and makes it possible to audit whether gates are actually holding.
6. Track conditional passes to closure. Every "go with conditions" gets an owner and a date, and someone confirms the condition closed. Untracked conditions are how defects sneak past a gate that technically did its job.
7. Review whether the gates are working. Watch for gates that never fail anything, a sign the criteria are too soft or the gate is rubber-stamping, and tune the criteria so the gate catches what it should.

How are quality gates different from inspection?

Inspection checks a product against a spec. A quality gate checks a whole stage against a set of criteria and controls whether work advances. Inspection asks "is this part good?"; a gate asks "is this stage complete and correct enough to build the next stage on?" Inspection is usually one input to a gate, not the gate itself, a passed first article is evidence you bring to the validation-to-launch gate, but the gate also weighs the control plan, the process capability, and the training.

The bigger difference is authority over flow. An inspection result gets recorded; a gate result gates. That is also the difference between a gate and a review meeting: a review discusses status, a gate decides passage. In production planning terms, gates are how you keep a schedule honest, work does not count as done because a date passed, only because it cleared the gate.

Gates only work if the evidence they need is actually available at the gate. When deliverables, test results, and check data live on scattered clipboards and spreadsheets, gate reviews stall waiting for someone to assemble the evidence, and conditional passes go untracked because nobody can see what is still open. When the checks and records flow into one live layer, a gate review can see the real state of the work in the moment, and open conditions stay visible until they close. Harmony connects the checks operators already run and the systems around the line into one live operational layer with no rip-and-replace, so the evidence a gate needs is there when the gate convenes. When CLS moved its production and quality logging off paper the data it was already generating became visible in real time, the raw material a working gate depends on. Set the boundaries, write hard criteria, protect the escalation path, and no defect advances on a deadline alone.