Skip to content

Go/No-Go Gauge

Physical pass/fail tool — instantiates Tolerance Band Management

Turns a tolerance into a physical pass/fail check — one end must fit, the other must not — so conformance is decided in seconds without reading a number.

A Go/No-Go Gauge is a tolerance band made solid. Instead of measuring a feature and comparing the value to limits, it embodies the two limits directly: the "go" element is sized at the maximum-material limit and must pass, the "no-go" element is sized at the minimum-material limit and must not. A part that accepts the go and rejects the no-go is in band; anything else fails. Its distinguishing virtue is that it collapses measurement, comparison, and verdict into one physical act — no reading, no interpretation, no skill — which is why it lives at the line rather than in the lab.

Example

An airline gate agent needs to decide, hundreds of times a flight, whether a passenger's carry-on is within the cabin allowance. Measuring each bag with a tape and checking three dimensions against limits would be slow and arguable. Instead there is a bag sizer: a rigid frame built to the maximum allowed dimensions. The rule is pure go/no-go — if the bag drops into the frame, it flies; if it doesn't fit, it's gate-checked. No numbers are read, no judgment is exercised, and the outcome carries its own correction: a no-go bag is tagged and sent to the hold. The frame is the tolerance band, and it decides in one motion what a written spec plus a tape measure would take a minute and an argument to settle. The same logic runs a plug gauge on a machined hole: go pin enters, no-go pin doesn't, part passes.

How it works

Its distinguishing move is to materialize the limits so the check needs no measurement rule at all — the gauge is the comparison. Because a borderline part could pass or fail on gauge wear or a fraction of a micron, the gauge's own dimensions carry a guard band: they are made slightly inside the true tolerance and are themselves tightly toleranced, so the gauge errs toward rejecting marginal parts rather than passing bad ones.[1] And because the outcome is binary and physical, the correction is immediate and built in: a no-go result routes the part straight to rework, scrap, or quarantine, with no separate decision step. What the gauge deliberately gives up is information — it says pass or fail, never how far off.

Tuning parameters

  • Guard-band offset — how far inside the true limits the gauge is built. More offset almost never passes a bad part but starts failing good ones; set it from measurement uncertainty and the cost of an escape.
  • Gauge tolerance and wear allowance — how tightly the gauge itself is made and how often it's replaced; a worn go-gauge silently starts passing oversize parts.
  • Which limits are gauged — one-sided (only the critical limit) versus both; gauging only the dangerous direction is cheaper but blind to the other.
  • Correction routing — what a no-go triggers (rework, scrap, quarantine, re-gauge), matched to the part's value and repairability.
  • Feature coverage — a simple size gauge versus a functional gauge that checks size and geometry together as the mating part would see it.

When it helps, and when it misleads

Its strength is speed, repeatability, and de-skilling: anyone can run it, it can't drift in interpretation the way a human reading a dial can, and it renders a defensible pass/fail thousands of times an hour with a correction baked into the "no." Its cost is exactly its simplicity. It throws away the magnitude of deviation, so it can't tell a process drifting toward the edge from one comfortably centered — a run of just-barely-go parts looks identical to a run of perfect ones until the gauge suddenly starts rejecting. Its classic misuse is trusting a worn or uncalibrated gauge, which fails toward acceptance and passes bad parts invisibly. The guard is to calibrate and retire gauges on a schedule, and to pair go/no-go checking with a variables measurement somewhere upstream when you need to see drift coming, not just catch it after it arrives.

How it implements the components

  • tolerance_band — it physically embodies the upper and lower limits as the go and no-go elements; the gauge is the band.
  • guard_band — its dimensions sit conservatively inside the true limits so measurement uncertainty and wear bias it toward rejecting marginal parts.
  • corrective_action_rule — a no-go outcome directly triggers the correction (rework, scrap, quarantine); the verdict and the action are one step.

It does NOT author the band it embodies (that comes from Engineering Tolerance Specification), stay trustworthy on its own (it depends on Calibration Procedure), or reveal drift over time (that's Statistical Process Control Chart).

References

[1] A guard band is a conservative margin placed inside a formal limit to absorb measurement uncertainty and instrument wear, so borderline items are rejected rather than mistakenly accepted. In gauge design this shows up as the gauge maker's convention of building gauges to a fraction of the part tolerance and biased toward the material condition — the mechanism failing safe rather than failing permissive. ↩