The sample and the container are made by the same factory and under different conditions. The gap between them is where OEM projects lose money.

An OEM project begins with a quotation and ends with a container, and the quotation covers only what happens in between the two. It does not cover the sample that has to be made first, the second sample that has to be made after that, or the third.

The production cost is the part of an OEM project that everyone can see and argue about. It is also the part that is least likely to be the reason a project runs late or lands badly. The cost that is not quoted, and rarely planned for, sits in two places: a loop of samples travelling back and forth between buyer and factory, and a discontinuity inside the factory between the sample that was signed and the production run that followed it.

This article deals with those two points rather than with the steps of the process, which are set out separately in our overview of the full OEM process from sample to container. The distinction is deliberate: the sequence is easy to describe, and the two places where it fails are not.

A quotation for an OEM order is a statement about manufacturing. Its cost, its lead time and its minimum all describe the run itself, and none of them describe the work required before the run can start.

There is a rough arithmetic that is worth doing before starting. A single sampling round is typically several days to produce, several more to ship, and then however long the buyer takes to open the box and form a view. Two to five weeks per round is ordinary. Three rounds is not an unusual number, and it will frequently exceed the production lead time that follows it.

None of this is a criticism of the process. It is a description of where the money and the time actually go, and it suggests that a buyer planning an OEM project should plan the sampling as a project phase with its own duration rather than as a formality before the real work starts.

The usual explanation for a second sample is that the manufacturer did not follow the instruction. In practice the more common cause is the way a first sample is evaluated.

When a sample is opened, the most obvious difference announces itself first. It might be that the colour is slightly off, or the surface is shinier than expected, or the edge is not quite as clean. Attention lands on that difference, feedback is written about it, and the sample is despatched back for correction. The second sample corrects the first problem, and a second difference becomes visible, which the first sample also had and nobody noticed because something else was more obvious.

The remedy is to front-load the evaluation rather than accelerate it. Instead of asking whether the sample is right, which invites a yes or a no about the most obvious feature, the useful approach is to work through the full requirement list against the sample in one pass and record a value or an observation for every item, whether or not there appears to be a problem.

There is a second contributor to repeated rounds that is worth removing early. Several requirements cannot be judged from a small sample at all, because they relate to how the material behaves over a run, how it handles, how it machines or how it looks in a large panel. Where those requirements matter, saying so before the sample is made allows the supplier to send something appropriate, rather than discovering after the fact that the sample was never able to answer the question.

A sample is evidence about capability and is not evidence about consistency, and confusing the two is the origin of the second problem.

This is not an argument against sampling. Sampling is the only practical way to establish that both parties mean the same thing before a container is made. It is an argument for being clear about what has been established, so that the gap between capability and consistency is closed by something else.

The subject of variation specifically, and why consecutive production runs from the same line differ, is covered in our piece on colour and batch consistency, which deals with the mechanisms that make a second container different from a first.

Between signing a sample and starting a production run, something changes inside the factory, and it is not a change anyone writes down.

A sample is a short event. A production run is a long one, and extrusion behaves differently over the two. A line that has just started is not at the same thermal equilibrium as a line that has been running for hours. Colour mixed for a short run behaves differently from colour fed continuously into a long one. Cooling conditions differ between a handful of sheets and a stack coming off the line continuously. None of these are faults; they are consequences of operating a line for different durations.

The production variables themselves, and how they determine whether a sheet comes off clean or flawed, are set out in our guide to extrusion variables in foam board production. The point for an OEM buyer is narrower: the variables that matter in a run are not the same ones that mattered while the sample was being made.

There is one question that determines whether a sample is meaningful evidence, and it is almost never asked at the point where it would do any good.

The question is easy to ask and easy to answer, and it belongs in the first conversation rather than after the sample has been signed. Asking it also signals something useful to the supplier: that the buyer understands the distinction between capability and consistency, and will be evaluating the sample on the basis that matters.

The third question concerns who made it. Where a factory runs multiple lines, a sample produced on one line and a production run scheduled on another are two different claims, and the capability of the second has not been demonstrated at all. The same applies where a sample is made on a pilot line, which is a legitimate way to develop a product and not a demonstration of production capability.

The capability of a factory to do this kind of work at all, and how to verify it before committing to an order, is the subject of our guide to vetting a board manufacturer, which covers what to look at beyond the sample itself.

The discontinuity cannot be eliminated, because a sample and a run are genuinely different operations. It can be caught early, and catching it early is cheap.

Step two connects this to the wider discipline of specifying measurable properties rather than descriptions. A sample whose colour has been measured and recorded can be compared objectively against a later run; a sample that exists only as a physical object cannot be compared against anything once it has faded or gone missing.

The tolerance framework that makes this possible, including how to state dimensional requirements so that a later measurement means something, is set out in our guide to dimensional tolerance and acceptance criteria.

An OEM order is an agreement about a specification, and the specification is only enforceable if it is written in a form that can be checked at both ends.

Where a first OEM project is being planned, the useful sequence is to write the specification with measurable values before any sample is requested, to ask the line and position questions at the first enquiry, and to plan for the sampling phase to take several weeks rather than several days.

The pattern that runs through this whole group applies here as clearly as anywhere. A specification should be a set of values that can be measured, and an order should be an agreement about how those values will be checked. A sample is a useful step in establishing that agreement and a poor substitute for it, because a sample made once says nothing about what will be made a thousand times.