Every inspection system eventually has to write a verdict in a box. The interesting design question is not how to decide Pass or Fail — that is a comparison — but what to write when neither is true.
The binary trap
Suppose a report has a Result column that can only say Pass or Fail. Now consider the rows that genuinely arrive in a real inspection:
- A row that was measured and is inside its limits.
- A row that was measured and is outside its limits.
- A row that was measured, but for which nobody ever specified limits.
- A row on the checklist that nobody measured at all — the feature was obscured, the operator was interrupted, the part moved.
Only the first two have an honest binary answer. For the other two, a two-state column must invent one. And it will almost always invent Pass, because the comparison "is the value outside the limits?" is false when there is no value and false when there are no limits.
The result is a report where a row that was never measured is indistinguishable from a row that was measured and passed. That is not a cosmetic problem. It is a document that asserts something nobody checked, over an operator's name, and it will hold up perfectly well right up until the moment it matters.
An inspection record should never be able to claim more confidence than the inspection actually produced. If a row cannot be judged, the report should say it cannot be judged.
Four honest states
Measuret's tolerance evaluation has four outcomes, and each corresponds to a genuinely different situation.
| Result | Value? | Limits? | Meaning |
|---|---|---|---|
| Pass | Yes | Yes | Measured, and within the inclusive limits. |
| Fail | Yes | Yes | Measured, and outside them. Printed bold red in both Excel and PDF. |
| Indeterminate | Sometimes | Sometimes | Measured but untoleranced, or toleranced but never measured. Something is missing, so no verdict is possible. |
| Not specified | Yes | No | No tolerance was ever declared for this row — it is a recorded observation, not a judgement. |
Indeterminate is the important one. It is not a failure of the software or an error state. It is the software declining to answer a question it was not given enough information to answer, and saying so in the box where a reader will see it.
Reading a report, the states tell you different things about what to do next. A Fail is a part problem. An Indeterminate is a process problem — either the checklist is incomplete or the inspection was. Both need attention; they need different attention.
Margin, and the guard band question
Pass and Fail describe which side of a line a value fell on. They say nothing about how close it came, and that distinction drives most real quality decisions.
A feature limited to 5.90–6.10 mm reading 6.09 mm passes. So does one reading 6.00 mm. Treating those as the same result throws away the most useful signal you have: the first is a process about to start failing, and you would rather know now.
This is why Measuret offers a Margin column — how much room was left before the nearest limit. A run of reports whose margins are shrinking is a drifting process, visible several batches before the first Fail.
There is a harder question behind this, and it is worth being explicit that software cannot settle it. If your measurement carries an uncertainty of ±0.02 mm and a value lands 0.01 mm inside the limit, the value passed but you cannot honestly claim the part did. Some quality systems require a guard band — tightening the acceptance limits by the measurement uncertainty — and some accept the reading as it stands. That is a decision for your quality system and your customer.
What software can do is make sure the facts needed for that decision are all on the sheet: the value, both limits, the margin and the unit.
The unmeasured row
A pattern is an ordered checklist of the measurements a part is supposed to have. The question of what to do when a row is left empty is more consequential than it sounds, because the empty row is the one that carries information.
Measuret gives you the choice explicitly, through the Unmeasured pattern rows option in Report Settings:
- Include them and they print as Not measured, so the report shows what was expected as well as what was found. A reader can see immediately that six of eight rows were completed.
- Drop them and the report lists only what was actually measured.
For anything that will be read by a customer or an auditor, including them is almost always right. A report listing six results looks complete. A report listing eight rows of which two say Not measured is honest, and invites the question that ought to be asked.
How this works in the application
A pattern is imported from an Excel workbook, or authored in the application. Only the description is required; the type, colour and limits are optional.
| Field | Effect |
|---|---|
| Description | Printed row description, and the legend drawn beside the measurement on the image. |
| Type | Linear, Angular, Radius or Any. A tool of the wrong kind is refused with feedback rather than silently mis-bound. |
| Colour | The overlay colour, and the painted Colour cell in the results table — so table and image match by eye. |
| Min / Max | Inclusive limits. A row with neither reports as measured rather than Pass or Fail. |
In the checklist panel each row shows its state as Pending, Armed or Measured. You arm a row, take the measurement, and it is recorded against that row before the list advances to the next one. Re-measuring an armed row replaces its result rather than adding a second, so there is exactly one value per row and no ambiguity about which one counts.
Pass, Fail and Indeterminate are then evaluated when the report is produced, from the value and the limits on each row. Rows that were never measured project as missing and print as Not measured if you have chosen to include them.
Without a pattern at all, the report simply lists whatever was authored, in the order it was authored, untoleranced — which is the honest description of what an untoleranced measurement session produced.
Whether a borderline result may be accepted, and whether measurement uncertainty must be subtracted from the tolerance before judging, are decisions for your quality system and your customer — not for the software. Measuret reports the value, the limits and the margin so that decision can be made with the facts in view.