I can't travel to China for the machine test — can an independent engineer attend for me?

The question is almost always asked late — the machine is nearly finished, the factory has proposed a date, and the flights, the visa or the project simply do not allow it. The honest answer is that delegation works well, and that the thing being delegated is not attendance. It is judgement.

What “attending for you” has to mean

A general inspection visit and technical representation are both legitimate services, and they are strongest at opposite ends of the problem. Lot-based inspection is built for sampled batches: quantity, workmanship, labelling, dimensions, packing, basic conformance against a checklist. That is the right instrument for repeat goods moving in volume. A machine test is a different object. Nobody is sampling a population; one unit either is or is not under control, and the evidence lives in how it behaves rather than in how it looks.

The practical difference shows up in who owns the agenda. A visitor witnesses the factory’s programme. A representative sets the test back against the approved protocol, and does five specific things before anything else happens:

• Confirms identity before the run — serial numbers and nameplates checked and photographed, so the record belongs to the unit you are buying and not to a demonstration machine that happens to be available.
• Forces the test back to the approved protocol — the agreed sequence, durations and conditions, not a shorter programme improvised for the visit.
• Records actual measured values — numbers with units and instrument references, rather than a column of ticks next to the word “pass”.
• Treats interventions as evidence — resets, tuning changes, alarm clears and operator adjustments logged when they happen. These are the most information-rich events of the day and the ones most likely to disappear from a summary.
• Locks the release rule before the result — what closes, what blocks shipment and what carries to site, decided from the protocol rather than negotiated once everyone is tired and the container is booked.

Everything on that list also appears in the factory acceptance test checklist — the agreed test plan, measured values, equipment identity, deviations and release rules locked down before shipment approval. If you are briefing a representative, that checklist is a reasonable place to start the conversation.

What if the factory manages what I see on the video call?

This is the right worry, and the answer is to control the evidence rather than the camera. Remote witnessing is well established in industrial procurement, and the structured versions of it depend on the same things a physical visit does: a defined protocol, controlled access to what is being shown, and a record that a third party can re-read. In practice that means several camera angles rather than one, a portable close-up view available on request, live document sharing so revisions can be confirmed on screen, and traceable equipment identification before the run rather than after.

It also means writing down what was notshown. A refused close-up, a camera that moved away during an alarm, a panel that was not opened — these belong in the record. A remote witness who reports only what the supplier chose to display has documented the supplier’s presentation, not the machine.

How do I know the person on the floor understands the machine?

A title does not prove competence, and this is the failure mode that costs most because it is invisible until commissioning. The international framework for credible inspection, ISO/IEC 17020, organises the question around three things: defined competence requirements for personnel, impartiality, and protection of the work from commercial pressure. It is a useful lens for a buyer even outside a formal accreditation context — if the person attending your machine test cannot independently interpret what they are seeing, their physical presence does not reduce your risk.

Tooling and moulding make the point sharply, because trials are diagnostic events rather than performances. A T1 trial is the first time the finished tool runs in a press, to prove that it fills, packs and ejects and to expose what still needs correction; T2 follows tooling changes or process tuning and shows whether the programme is converging. Process truth in moulding is created inside the mould, not on the machine controller display — which is why cavity-pressure behaviour is used to diagnose faults such as short shots and burn marks that a controller screen will happily report as normal.

A representative with that background challenges switchover, hold pressure, venting, cooling balance, ejection behaviour and the stability of the process window. The question is not whether the machine produced a few acceptable pieces on the day. It is whether the tool and the machine are actually under control.

What if it passes for twenty minutes and drifts when the run gets real?

Then it has not passed. Repeatability and endurance are the point of a machine acceptance test, which is why endurance runs of several hours at production conditions are a normal requirement in machine and automation contracts — and why they are the requirement most often quietly shortened when the schedule tightens. A machine that looks clean for a short demonstration and degrades with heat, cycling, pressure variation or operator intervention is still a commercial risk; it has simply moved the discovery to your site.

A representative holds the factory to the agreed duration, conditions and measurable criteria, and records pressure, speed, temperature, vibration, alarms, interruptions and every supplier adjustment made during the run. Mid-run tweaking is treated as a control issue rather than a minor note, because a process that needs continuous intervention to stay inside tolerance will not survive being staffed differently at your plant.

On moulding trials the same logic becomes a question about the process window — the range of temperature, pressure and time within which acceptable parts can actually be produced. If the factory can only make good cycles at knife-edge settings, you are not watching production readiness. You are watching a fragile demonstration that will fail once the machine is installed, staffed by someone else and run for volume.

What proof do I get afterward, so I am not trusting a summary?

A decision-grade record, not a travel note. The package should tie the exact unit to the agreed procedure and let someone who was not in the room reconstruct what happened:

The practical standard is simple to state. If the post-test package would not survive a supplier dispute, an internal gate review or a payment-hold decision, it is not enough. What a good representative leaves you with is traceable evidence you can act on commercially, technically and contractually — not confidence that somebody was physically present.

What to send when you delegate

Delegation goes wrong most often at the briefing, not on the day. The minimum useful pack is short:

• The approved test protocol — or the supplier’s draft, which is usually where the shortened durations and vague acceptance wording are already visible.
• The specification and drawings the machine was bought against — including any deviation conceded during negotiation, because those are the clauses most likely to be quietly re-interpreted.
• The acceptance criteria that actually matter to you — cycle time, output rate, tolerance, uptime, changeover, in your own words if the contract is vague.
• The release rule— what the FAT result is tied to commercially: a payment milestone, shipment authorisation, or nothing at all.

For the full running order on the day — readiness, document requests, identity checks, functional witnessing, non-conformities and the release decision — see the factory acceptance testing guide. For the boundary between what the factory test can prove and what only site commissioning can, see FAT vs SAT.

Frequently asked questions