Panel X-Out Allowance: The Quantity You Ordered vs the Units You Can Use

Two numbers both called "quantity"

You order 500 pieces of an HDI board. The supplier quotes and ships in array format, say a panel carrying a number of identical units. Your purchase order says 500. The packing list says 500. The invoice says 500. And yet the line that was supposed to build 500 assemblies stops short.

Nothing dishonest happened. What happened is that "quantity" was never defined, and in array-format delivery there are two candidates:

• Panel count — how many panels leave the fab. This is what the fab plans, prices and ships against. It is a production unit.
• Usable unit count — how many individual boards on those panels are actually good and not marked as scrapped. This is what your assembly line consumes. It is a demand unit.

Under a perfect yield these two numbers are related by a fixed multiplier and nobody notices the distinction. The moment a single unit inside a panel fails electrical test, the multiplier breaks. The fab marks that unit — an X-out, sometimes called a scrap mark or a cross-out — and ships the panel anyway, because the rest of the panel is good and scrapping the whole array over one unit would be wasteful for both sides.

That practice is normal and, in most cases, in your interest. The problem is not that X-outs exist. The problem is that the gap between panel count and usable unit count has to be assigned to someone, and a PO that states only one number assigns it by accident.

If your PO reads "500 pcs" and the contract does not define what a piece is, you have left the interpretation to whoever is reading it later. The fab reads it as a production target it can meet. You read it as a build quantity you can consume. Both readings are defensible. Only one of them gets you boards.

Is an X-out a normal output or a quality problem?

This is the question that decides the tone of the whole conversation, and buyers and fabs often answer it differently without realising they disagree.

The fab's position is that an X-out is evidence the process worked. Electrical test found a defect, the unit was marked so it cannot be assembled by mistake, and the defect was contained before it reached you. A panel with a marked unit is a panel that was inspected. A fab that never showed you an X-out would be a fab whose test coverage you should question.

The buyer's position is that an X-out is a unit that was paid for and cannot be used. Whether the fab's process caught it is procedurally reassuring and commercially irrelevant.

Both are correct, which is why this cannot be settled by argument. It has to be settled by a term. The allowance percentage is that term: it draws a line and says that below it, X-outs are normal process output absorbed by the agreed price, and above it, they are a shortfall the supplier makes good.

Two consequences follow immediately, and both are routinely missed:

1. An allowance is a cap, not a forecast. Agreeing to an allowance does not mean you expect that many X-outs, and it does not entitle the supplier to deliver up to it as a matter of course. It defines where the commercial boundary sits. Treating the cap as a planning number is how buyers end up chronically short.
2. An allowance without a remedy is decorative. "Up to X percent X-outs acceptable" tells you what is tolerated. It does not say what happens when the limit is exceeded — replacement at the supplier's cost, credit, partial reship, or renegotiation of the next lot. If the remedy is unstated, exceeding the allowance produces a dispute rather than a resolution.

Before you negotiate the number, decide which of those two readings your own planning assumes. If your material requirement planning consumes usable units but your PO commits panels, you have an internal mismatch that no supplier term can fix. The broader framework for stating what you will accept on arrival is covered in incoming inspection and acceptance plan for PCB delivery; this article deals with the one line in it that is about counting rather than quality.

The three terms buyers forget to put beside the percentage

Most X-out negotiations stop at a single number. The number is necessary and insufficient. Three further terms determine whether that number protects you, and all three are cheap to agree at RFQ stage and expensive to argue after delivery.

1. How an X-out is marked. A marked unit has to be unmistakable to three different audiences: your incoming inspection, your assembly operators, and any automated handling in between. A marking method that satisfies one can fail the others. Ink on the solder side is visible to a human and invisible to a top-side camera. A label is visible to everyone and may interfere with placement or reflow. A notch or drilled hole is permanent and may not survive your own depaneling plan. Routed-out removal leaves an unmistakable void but changes the panel's mechanical behaviour in handling and in the reflow oven.

The point is not that one method is correct. The point is that the method has to be named in the PO, because the right choice depends on your downstream process, not the fab's convenience. If you assemble the panel yourself, the marking must be detectable by whatever decides where to place parts. If you depanel first and assemble loose boards, the marking only has to survive until singulation. Ask for one method, in writing, and ask whether that method is what the fab normally applies — a non-standard marking request can carry a cost or a process constraint, and you want to hear that before the order, not after.

2. Distribution limits, not just a total. An allowance expressed purely as a percentage of total units is blind to where the bad units sit. Two deliveries can satisfy the same percentage and behave completely differently:

• X-outs spread thinly across many panels. Every panel is usable, each one yields slightly fewer assemblies, and your line never stops.
• The same total concentrated in a few panels. Those panels may be effectively unusable for your process, while the percentage remains compliant.

Concentration matters most if you assemble in array format, because a panel below a usable-unit threshold may not be worth setting up at all. If that is your situation, state a per-panel limit alongside the overall allowance, and state whether adjacent X-outs are permitted. Two scrapped units side by side can break panel rigidity or interfere with support tooling in ways that two separated units do not.

3. Whether partial panels count. If an allowance is exceeded and the supplier reships, you need to have agreed in advance whether a partial panel — fewer units than a full array, perhaps a cut-down panel — is an acceptable form of make-good. For a buyer who depanels on arrival, usually yes. For a buyer whose assembly fixtures are built around a full array, a partial panel can be worthless. This is a one-line term that prevents a reship from arriving in a form you cannot run.

None of these three require you to specify numbers you do not have. They require you to state which of your own processes consumes the panel. If you are not sure which assembly route a given build will take, say so in the RFQ and ask the supplier to quote the marking method and distribution limit for both routes.

What an X-out actually costs you when you assemble the panel yourself

The board price is the smallest part of this. If you hand the panel to a contract assembler as a turnkey item, the X-out shows up as a material shortfall and gets resolved with the fab. If you assemble it yourself, or you supply the panels to an assembler as consigned material, the scrapped unit costs you money in three places that no quotation lists.

Feeder and material consumption. A placement program either skips the marked unit or it does not. If the machine has no bad-mark sensing configured for that product, or the marking method cannot be read, the program places a full set of components onto a unit that will be thrown away. Every component on that unit is lost, including anything long-lead or single-sourced. On a dense board this can be the largest single cost in the whole X-out conversation, and it is entirely invisible until someone counts the kit at the end of the run.

Program and setup effort. If the machine can skip marked units, somebody has to configure that: the bad-mark position has to be defined, the sensing has to be verified, and on the first panel of a new lot it has to be checked rather than assumed. That is engineering time per lot, not per order. If the marking position moves between lots because the fab marks wherever the defect is, the verification cost repeats. This is the practical reason to ask for a *consistent* marking location, not just a marking method — and it connects directly to how your placement data is prepared, which is covered in CPL and fiducial requirements for PCB assembly.

Thermal and handling behaviour. A panel with routed-out units does not behave like a full panel. Support requirements change, and so can warpage behaviour through reflow. A panel with several units removed in the same region is a different mechanical object from the one your process was qualified on. If your allowance permits concentrated or adjacent X-outs, you have implicitly permitted this, which is the second reason distribution limits belong in the PO.

There is a fourth cost that is harder to price and worth naming anyway: planning noise. When the usable unit count per delivery is variable and unannounced, every build becomes a small reconciliation exercise. The fix is not a tighter allowance. The fix is for the supplier to state the usable unit count per shipment on the packing documentation, so the number you plan against arrives with the goods instead of being discovered on the line.

Ask for that. It costs the supplier nothing — the count already exists, because the units were marked deliberately — and it converts an end-of-run surprise into a receiving-dock fact.

How to count and how to prove it

Counting X-outs is simple until there is a disagreement, at which point everything depends on what was recorded and when.

Count at receiving, not at the line. If panels are opened, split across storage locations, partially consumed and then found short, you can no longer prove the condition they arrived in. Receiving is the last point where the delivery exists as a single identifiable lot. This is the same logic that applies to any arrival-condition claim, and it is why the acceptance plan and the count method should be one document rather than two.

Record three numbers per shipment, not one. Panels received, usable units, marked units. One number cannot support a claim; three can, because they reconcile. If the three do not add up, that is itself a finding worth raising before the goods are consumed.

Photograph marked panels before singulation. After depaneling, a marked unit becomes an unmarked loose board that failed, and the evidence that it arrived marked is gone. A photograph of the panel with its identification visible is the entire proof in most disputes. This is not an accusation of bad faith; it is the ordinary consequence of the marking being attached to a physical object you are about to destroy.

Tie the count to a shipment identifier the supplier also holds. A count that references your internal receipt number only is a count the supplier cannot verify. Reference the panel lot or shipment identifier that appears on their documentation.

Where to raise the result depends on what you find. A count within the allowance is a record, not a complaint; it belongs in supplier performance data and informs the next negotiation. A count above the allowance is a claim, and it should cite the agreed allowance, the agreed remedy, and the three reconciling numbers. Sending the claim without the agreed remedy in it invites a renegotiation you already had.

Two ways to write the quantity line

The difference between these two PO formulations is the whole subject of this article.

Formulation A — panel-committed. The order states a panel count. X-out allowance, marking method and distribution limits are stated separately. You are buying production units and accepting defined variability in usable output. This is appropriate when you depanel on arrival, when your assembly volumes have buffer, or when you are price-sensitive and willing to absorb variance in exchange for not paying for the fab's protection against it.

Formulation B — usable-unit-committed. The order states a minimum usable unit count. The fab decides how many panels to run to deliver it, and the X-out risk is theirs. You will pay for that transfer, directly or through yield allowance built into the price, and lead time may carry a margin for a possible second run. This is appropriate when a shortfall stops a line, when the assembly is high-value enough that running short costs more than the premium, or when you cannot absorb planning variance.

Neither is the correct answer. What is incorrect is writing Formulation A and planning as though you had agreed Formulation B, which is the default failure mode because A is what the industry quotes in and B is what production planning consumes.

If you are comparing suppliers, check which formulation each quotation assumes before you compare the prices, because a Formulation B quote will look more expensive than a Formulation A quote for reasons that have nothing to do with the supplier's cost base. The same distinction shows up in how array format drives unit price in the first place, which is covered separately in HDI PCB panel utilisation, array format and cost in the RFQ.

What to put in the RFQ

Six lines. None of them require a number you do not already have.

1. Quantity basis — panel count or minimum usable unit count. State which.
2. X-out allowance and remedy — the limit, and what happens above it.
3. Marking method — which method, and that it must be consistent across the lot.
4. Distribution limit — per-panel maximum, and whether adjacent X-outs are permitted.
5. Partial panels — acceptable as make-good, or not.
6. Documentation — usable unit count stated per shipment on the packing documentation.

Ask the supplier to answer all six in the quotation rather than in a follow-up email, so that the answers carry the same commercial weight as the price. If you want help structuring the quantity and acceptance section of an array-format enquiry, send the enquiry through PCB168 and state which of the two formulations you need; the construction options that can be quoted are listed on the HDI PCB product page and on the capabilities overview, and those represent quotable ranges rather than a commitment for any specific order.

FAQ

Is a panel with X-outs a defective delivery?

Not by itself. Within an agreed allowance, marked units are normal process output and the delivery conforms. Without an agreed allowance, there is no standard that makes the delivery either conforming or defective, which is why the term matters more than the argument.

Should I just ask for zero X-outs?

You can, and you will be quoted for it. Zero means the fab absorbs all scrap by running more panels, and that cost returns in the price, the lead time, or both. It is a reasonable request for small high-value builds and an expensive one for volume orders.

Does the allowance apply per shipment or per order?

Whichever you write. If you do not write it, a multi-shipment order can satisfy an order-level allowance while an individual shipment leaves you short. If any single shipment feeds a build, apply the limit per shipment.

Who pays for components placed on a marked unit?

Whoever owns the material at that point, which means it falls on you under consignment and on the assembler under turnkey — unless the marking was unreadable or inconsistent with what was agreed, in which case it becomes a supplier issue. This is exactly why the marking method belongs in the PO.

Can I reject the whole panel if one unit is marked?

Only if the contract gives you that right, and few do, because it transfers the full scrap cost of the array to the supplier over a single unit. If a panel below a usable-unit threshold is worthless to you, express that as a per-panel distribution limit instead. It gets you the same protection without asking the supplier to price a liability they will refuse.

Do X-outs affect my own test and traceability records?

Yes, and in a way worth planning for. A marked unit that is never assembled still occupies a position in your panel map, so position-based traceability has to record it as absent rather than as a missing record. Deciding this before the first lot avoids a reconciliation problem later.