Four of the five factory acceptance tests in the delivered register are witness points [TARGET]. Factory acceptance is weighted in the commissioning index at 0.15, the same as site acceptance, on one stated ground: it is the only window that contains the characteristics flagged site-invisible. A witness point is the grade that lets the work continue if the named party does not attend. Those two numbers do not describe the same arrangement. Where the answer to the mandatory factory field is no — the characteristic cannot still be verified once the equipment is installed — the factory test is the only time window, a missing record there fails a hard gate, and the field definition itself states that the gate may not be closed by looking again at site acceptance [FACT].
Hold and Witness Differ by the Consequence of Absence
Both symbols are defined in the procurement package, and so is what separates them. At a hold point no following operation may proceed without written release by the named party; an absence stops the work and cannot be signed off afterwards. At a witness point the named party has the right to attend, and if it does not attend within the agreed notice period the work continues, with the evidence retained in full. Attached to the pair is the operative sentence: the difference between them is not importance but the consequence of absence, and anything that cannot be remedied after an absence has to be set as a hold point.
Site-invisible characteristics are described in the commissioning index in exactly those terms. The examples it gives are the earthing arrangement at the terminals inside assembled switchgear, earthing internal to equipment, the earth grid, buried services and firestopping. Retained evidence has not been enough to restore any of them once a termination has covered them.
Two Site-Invisible Examples Sit on Witness, Not Hold
Thirty points sit in the register, one per test: seven witness points and twenty-three hold points, with all seven witness rows graded standard risk and the hold rows split twelve high and eleven critical [TARGET]. Five of the thirty are factory acceptance tests. Four of those five are witness points, and the only factory hold point is the generator controls and paralleling test, which is also the only factory test whose release authority names the utility. Every row remains open.
Weighting makes the window explicit. Design requirement and enquiry clause carry 0.05 each in the maturity calculation, receiving carries 0.05, installation and pre-commissioning 0.10 each, and factory acceptance, site acceptance and system performance testing 0.15 each, with the integrated test at 0.20 [TARGET]. Factory acceptance earns its 0.15 on one stated ground: it is the only window that contains the site-invisible characteristics. Nothing later in the chain can re-earn it.
Two of the index's own site-invisible examples land on tests the register grades as witness points: the switchgear factory test, which is where terminal earthing inside the assembly is seen, and the grounding and bonding continuity check. Firestopping and penetration inspection, the third example, is a hold point. Two of three therefore sit on the wrong side of the line the discipline sentence draws, and neither of the two is graded above standard risk.
The Package Sets Two Point Types; the Register Records One
Factory level is not set at a single point type by the procurement package. Key packages are set at witness and the check before crating at hold, which is two point types on one test. Each test in the delivered register has one point type, so it can express one of the two, and where it chose, it chose the half that lets the work continue. A register column carrying the site-invisible flag beside the point type would make the conflict visible row by row and would leave the point types alone; no such column exists, and the index's mandatory field is reproduced nowhere in the delivered set.
That decides what goes into the enquiry rather than what happens on a test floor. A witness-point plan is an enquiry deliverable, and the hold before crating has to be written into it before any equipment is built, because the characteristics at issue disappear behind their own terminations at the first termination.
One dependency is written identically on all thirty rows: the test-specific physical limit set approved, and six prefunctional checks closed. That limit set is a placeholder in the acceptance file for every one of the thirty. The delivered closed set contains three prefunctional checks.
Limits and open items
Confirmed: a factory item flagged as unverifiable after installation is a unique time window; a missing record there fails a hard gate and may not be closed at site acceptance. [FACT]
Register figures, not a test campaign: thirty points, seven witness and twenty-three hold; four of five factory tests graded as witness; hold rows split twelve high and eleven critical; chain weights 0.05 / 0.10 / 0.15 / 0.20; six prefunctional checks named against three that exist. Every row is open. No object yet carries the site-invisible flag, because no object exists. [TARGET]
Open: a witness-point plan for any package; the hold before crating written into that plan; a register column for the site-invisible flag; the physical limit set behind each row; notice periods no supplier has accepted. Nothing here has been executed. Nothing releases a crate. [HOLD]
Ask Which FAT Row Remains Observable After Termination
The useful test is not whether the factory visit is important. Ask which of the five factory rows would still be observable after the first termination. If the answer is "none of the four witness rows," the register has already chosen the half that lets the work continue.
© 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., “FAT Items Marked Site-Invisible Cannot Wait for SAT,” https://www.kkdatasvc.com/lab/rfq-vendor-review-and-commissioning/fat-items-marked-site-invisible-cannot-wait-for-sat/.
Discuss This Constraint. If your project record shows a different result, or the same failure domain under another name, we want to see it. Email inquiry@kkdatasvc.com or use the contact page.