Missing Criteria Needs Its Own Root-Cause Class, Not the Commissioning One

A test can run on calibrated instruments and still score what an unwritten test scores. The commissioning index writes that outcome as a seventh root-cause class [TARGET]. The class sits beside design and beside commissioning, not inside either. Responsibility runs to the design, record-engineer and owner decision chain. The corrective action is to obtain the criterion. Repeating the test does not close it. Without the class, the index says, these defects get filed as commissioning execution problems or absorbed by recording without adjudication.

The Ledger Already Lists Criteria-Missing as a Class

The defect record lists seven mandatory root-cause classes: design, procurement, manufacture, transport, installation, commissioning, and criteria-missing [TARGET]. The last one is not a leftover bucket. The same index concludes that a large share of tests on this project will approach the execution window with the criterion still unfrozen, and that folding those defects into commissioning is the error the class exists to prevent.

A template defect in that file is already filled as criteria-missing, severity two, disposition defer. The owner-accept reference is marked required but absent. The note says the record is a template and has not been executed. That record cannot close on its own terms. The class is therefore not a future procedure. It is a column the ledger has already drawn.

The index's only legal deliverable at this stage is structure. Filling a numeric criterion now would dress an unexecuted procedure as measured evidence without a criterion.

108 of 151 Acceptance Rows Still Have No Criterion

A delivered acceptance parameter file carries 151 rows across 30 tests [TARGET]. Of those, 108 carry a to-be-confirmed target and a to-be-confirmed tolerance and are marked not approved; 121 carry a hold identifier; and 43 are internally authored [TARGET]. None of the 151 has been accepted.

Forty-three internally authored rows split in two. Thirty are state-sequence acceptances whose target is a boolean true. The other thirteen are the only physical thresholds that carry a number.

Sixty-one rows are physical thresholds, and thirteen of the sixty-one carry a number at all [TARGET]. Those thirteen are a facility power baseline of 15 MW [TARGET], a billable computing baseline of 12 MW [TARGET], an efficiency ratio baseline of 1.20 [TARGET], an identity-match of 100 percent [TARGET], and nine zeros covering damage count, backfeed current, unintended output, unintended closing, reverse power, unintended discharge, hash mismatch and open critical defects. Every one of the thirteen is sourced to the project's own zero-event or design baseline and is still marked internal draft. Each of those thirteen still carries a hold identifier. The remaining forty-eight thresholds repeat one closure sentence: the authority returns the accepted value in the stated unit, with a test tolerance and controlled evidence.

Thirteen carry a value. The other 138 carry none.

The commissioning design package that publishes the file stops at ready for review by the commissioning authority. No discipline authority has read it. A to-be-confirmed limit is an execution hold. It is not acceptance flexibility.

A Test Without an Accepted Criterion Scores Zero

Evidence measured without an accepted criterion scores 0.00 [TARGET] in the index. A judged and accepted result scores 1.00 [TARGET]. A correctly executed test, on calibrated instruments, producing clean data, therefore scores what an unwritten test scores.

The execution state machine names the same situation twice. A record whose criterion source is to-be-confirmed is blocked and may not be executed. A blocked test executed anyway under written owner approval becomes executed but unadjudicated. A record judged as passed while its criterion source is empty fails the seventh hard gate outright.

Making that state visible has a stated reason. If the ledger has no such state, the commissioning party writes that the result accords with engineering practice, or simply ticks the box.

Forty-four external decisions are listed by the same index across six responsible parties, with none of them obtained [TARGET]. Two of those forty-four are first-priority letters, not self-sufficient unlocks. The other forty-two cannot independently unlock any joint scene, however far they are driven.

Obtaining a Missing Criterion Is Not a Retest

Any deviation observed during a test is registered as an anomaly first. The commissioning authority and the record engineer then qualify it into one of five classes, of which criteria-missing is one [TARGET]. The test is barred from closing until the qualification is written. Retest scope follows from the qualification and not from the symptom. Changing hardware, control logic, an upstream supply or the earthing reference forces a whole-system retest. Changing a design requirement forces the full chain from the first level. Obtaining a missing criterion does neither.

Which is why the class cannot sit under commissioning. Zero lands in the commissioning column while the cause sits in a decision chain that has produced nothing. The correction is a letter, not a retest.

Limits and open items

Confirmed: criteria-missing is a root-cause class of its own, separate from design and from commissioning; its responsible party is the design, record-engineer and owner decision chain; its corrective action is to obtain the criterion rather than to repeat the test; measured evidence without an accepted criterion scores zero and must not be judged as a pass. [FACT]

Register figures, not a campaign: seven classes; 151 rows, 108 empty and unapproved, 121 on hold, 43 internally authored; thirteen of sixty-one physical thresholds carrying a number, including 15 MW, 12 MW, 1.20 and 100 percent; five qualification classes; forty-four external decisions, none obtained. Every example card is a template and has not been executed. [TARGET]

Open: a commissioning authority to read the package; a named producer for the criteria-missing count the index wants on the first page of every gate review; any of the forty-four decisions returned; an accepted criterion on any of the 151 rows. [HOLD]

Nothing here has been reviewed or sealed by an Engineer of Record. Nothing describes how any installed system behaved.

Separation collapses if the owner rules that a criterion the project wrote for itself is acceptable evidence. Those thirteen internally sourced values show the ledger will already take a project baseline where no authority has answered, and it marks them internal draft instead of refusing them. Extend that and the missing criterion becomes a design output again, at which point the class it belongs to is design and the corrective action is a calculation rather than a request.

Who Produces the Criteria-Missing Gate-Review Count

The index asks for one figure on the first page of every gate review: the count of defects graded criteria-missing. The owner still needs to name who produces it.

© 2026 K&K Data Service Inc. All rights reserved. Reproduction or republication is permitted only with clear attribution to K&K Data Service Inc. and a working hyperlink to the canonical URL of this article. Excerpts must preserve the technical context, maturity labels, assumptions, and limitations. No excerpt may imply project approval, field validation, certification, or endorsement that the original article does not state.

Source: K&K Data Service Inc., “Missing Criteria Needs Its Own Root-Cause Class, Not the Commissioning One,” https://www.kkdatasvc.com/lab/rfq-vendor-review-and-commissioning/missing-criteria-needs-its-own-root-cause-class-not-the-commissioning-one/.