HDI PCB Supplier Shortlist: Rank by Evidence You Can Actually Get

Eight suppliers, eight yeses, and nothing to cut on

Shortlists stall for a reason that has nothing to do with capability. You send the same HDI enquiry to eight suppliers, and eight of them answer yes. Yes to the layer count, yes to the via structure, yes to the line width, yes to the schedule. Nothing in the replies separates them, so the shortlist gets cut on whatever is left: price, response speed, or who the sales contact reminds you of.

The replies look identical because you asked a question that cannot be answered differently. "Can you build this?" invites a yes. A more useful question is not about capability at all — it is about evidence: for each claim in that reply, when does it become possible for me to check it? Sort the answers that way and the eight replies stop looking alike, because suppliers differ enormously in what they can show you now versus what you only discover later.

The four tiers of a supplier claim

Every statement a supplier makes about an HDI job falls into one of four tiers, defined purely by the moment it first becomes checkable.

Tier 1 — checkable at quote. The supplier can produce something now, before you commit money, and you can read it yourself. Process ownership for each step, the scope boundary of their quality system registration, which test and inspection capabilities exist on site versus which are sent out, the stackup they intend to build, their own reading of your data package including the questions they would raise. These are statements about how the supplier is organised, so the evidence already exists and does not need your job to be built first.

Tier 2 — checkable at first article. The supplier's claim is real or not, but your specific job must be built once before anyone can tell. Registration behaviour on your layer arrangement, how your via structure behaves through their process, what their coupons and electrical test actually report on your netlist, whether the dimensional results on your panel land where they said. Nothing here can be resolved by a document, because the document would be about someone else's board.

Tier 3 — exposed only at volume. The claim is about consistency, not feasibility. Run-to-run variation, behaviour at larger panel quantities, what happens when the job is scheduled alongside other work, how they handle a lot that drifts. A first article cannot show you this — a first article is one lot, usually the one that got the most attention.

Tier 4 — not checkable during purchasing at all. Statements that are either unfalsifiable as written or whose evidence sits behind a boundary you have no access to: internal yield figures you cannot audit, priority you have been promised relative to other customers, the claim that a difficulty "will not be a problem," subcontractor arrangements you cannot inspect. These may well be true. You just have no mechanism to find out.

Why the tier matters more than the answer

Two suppliers both say they control your impedance target. One attaches the stackup they propose, names who owns the target, and states what their coupon reports and on which layers. The other says it is handled. Both answered yes. The first answer is Tier 1 — you can read it tonight and argue with it. The second is Tier 4 as written, and will be relabelled Tier 2 later when a first article either meets the target or does not.

That difference is the only thing in the reply you can act on before spending money. It is also, usefully, not a matter of opinion: either a document arrived or it did not.

A second consequence is about where risk lands. Tier 1 claims that turn out wrong get caught in the quote comparison, which costs you an email. Tier 2 claims that turn out wrong get caught at first article, which costs a build and a slot in your schedule. Tier 3 claims that turn out wrong surface mid-production, when switching supplier is expensive and slow. Tier 4 claims never get caught at all — they simply become whatever the contract says when something goes wrong, which is why their real home is the terms, not the shortlist.

Shortlisting is therefore the act of moving decisions out of the expensive tiers and into the cheap one. Any claim you can pull from Tier 4 into Tier 1 by asking for a specific artefact is a claim you have de-risked for the price of asking.

Working Tier 1: ask for artefacts, not confirmations

The practical move is to stop asking yes/no questions and start naming the artefact you expect back. The two questions below ask for the same information; only one of them produces evidence.

The right-hand column has two properties the left-hand column lacks. The answers differ between suppliers, which is the entire point of a shortlist. And a non-answer is itself informative: a supplier who will not state which steps are subcontracted has told you something real about the visibility you will have later.

Note the last row. You are asking what documentation normally accompanies a shipment — a question about their standard practice. You are not asking for another customer's files, which they should refuse to provide.

Pulling Tier 2 forward instead of hoping

Tier 2 claims cannot be converted into documents, because only your board can settle them. What you can do is decide — before the shortlist closes — which Tier 2 questions are worth paying to answer early, and on which boards.

Three things make that decision easier. First, ask each shortlisted supplier what they would want to prove on a first build and what they would report back; a supplier who already has a view on this is describing a process they run habitually. Second, agree what the first article report will contain before the build, not after, so you are comparing the same fields across suppliers rather than whatever each chose to measure. Third, separate "we need this verified" from "we need this verified by this supplier" — some Tier 2 items are properties of your design and will behave the same anywhere, and verifying those once is enough.

What you should not do is treat a successful first article as closure on Tier 3. One good lot demonstrates feasibility. It says nothing about variation, and reading it as a consistency guarantee is the most common way a shortlist decision quietly fails three months later.

Tier 3 belongs in the terms, Tier 4 belongs nowhere near the shortlist

Consistency claims cannot be screened, so stop trying. They are handled by what the agreement says happens when consistency fails: how a drifting lot is dispositioned, who decides, what evidence is retained and for how long, what the notification obligation is when something changes on their side. Those are contract questions, and a supplier's willingness to be specific about them is itself Tier 1 evidence — it is checkable now.

Tier 4 claims need a harder rule: a claim you cannot verify during purchasing must not be used as a reason to include or exclude anyone. This cuts in both directions. You should not shortlist a supplier because their internal figures sound better than the others', since you cannot check them. You also should not drop a supplier for declining to share something they are not obliged to share.

Left in the file, a Tier 4 claim does not sit still. It becomes an assumption in your plan, and the gap only surfaces where the plan meets reality: at incoming inspection when an attribute you assumed was covered turns out never to have been declared, at a change notification that never arrives because no one agreed it was owed, or in a failure investigation where the evidence you needed was not retained because retention was never specified. The claim itself was never the problem; building a schedule on it was.

Where the shortlist line goes

The line sits at the bottom of Tier 1: shortlist on what is checkable at quote, and carry everything below it as a known open item with an owner and a resolution point.

This sounds like it throws away information, and it does — deliberately. A shortlist built on Tier 1 evidence is a shortlist you can defend, reproduce, and hand to someone else. A shortlist built partly on Tier 4 impressions cannot be reconstructed even by the person who made it, which is why those decisions are so hard to revisit when the programme changes.

Carrying the open items is the other half, and skipping it is the usual failure. "Open" means written down with a name against it and a point in the schedule where it gets resolved — not a vague intention to check later. A Tier 2 item with no scheduled resolution point is simply a Tier 4 item you have not admitted to yet.

One request, all candidates, same format

A final mechanical point that decides whether any of this works: the evidence request must go to every candidate in the same form, at the same time, with the same response format.

Comparability is the entire value of a shortlist. If supplier A answered a long detailed enquiry and supplier B answered a short one, their replies are not comparable, and whatever you concluded is an artefact of your own inconsistency. The same applies to timing — a reply written under a deadline and one written at leisure differ in ways that have nothing to do with the supplier.

So: one request, one template, one deadline, one set of fields. Then rank by what came back, and record what did not. The suppliers who return Tier 1 artefacts without being chased twice have demonstrated something you will rely on for the entire programme, and they demonstrated it before you spent anything.

Related pages on PCB168

• HDI PCB manufacturing — the product page these enquiries usually start from
• Multilayer PCB production — when the job is better framed as multilayer than as HDI
• Production capabilities — scope overview for screening questions
• Request a quote — where to send the data package once the shortlist closes
• Contact — for engineering questions before the RFQ goes out
• HDI buyer hub: specify, compare, verify — the overview this article sits under
• Build class specification ownership — who decides the build class and when
• From application spec to buildable requirements — turning market language into quotable requirements
• RFQ qualification documents — which documents belong in a qualification pack
• Impedance target ownership — a worked example of a claim that needs an explicit owner
• Defining test and inspection requirements — writing test requirements that produce reports
• Incoming acceptance planning — where unverified claims surface first

FAQ

Is this just asking for more paperwork?

No — it is asking for less, but earlier and in a fixed format. The tier framework tells you which requests produce something you can compare and which produce agreement you cannot use. Requests that only generate a yes should be dropped, which usually shortens the enquiry.

What if a supplier refuses a Tier 1 request?

Distinguish two refusals. Declining to share another customer's files or genuinely confidential internal data is correct behaviour and should not count against them. Declining to state which process steps are subcontracted, or what the scope of their registration covers, is different: that information concerns your job and the visibility you will have, and the refusal is itself the answer.

Can a Tier 4 claim ever be moved up?

Often, by rewriting it as a request for an artefact. "Our yield is high" is Tier 4. "What is dispositioned as a reject at final inspection, and what report accompanies the shipment" is Tier 1 — same underlying subject, but now checkable. If no rewrite produces a checkable form, the claim belongs in the contract terms rather than the screening criteria.

Should a smaller supplier be penalised for having less documentation?

Not for volume of documentation, only for availability of the specific evidence your job needs. A small supplier who states plainly what they do in house, what they subcontract, and what reports you will receive has produced solid Tier 1 evidence. A large one with a thick brochure and no answer on your actual questions has not.

Does a passed first article mean the supplier is qualified?

It settles the Tier 2 questions you designed it to settle, and nothing else. Consistency is Tier 3 and is not observable in one lot. Read a first article as the feasibility answer it is, and handle consistency through the agreement.

How many suppliers should get the evidence request?

Enough that the replies are comparable and few enough that you will actually read them all. The binding constraint is usually your own review capacity: an unread reply provides no evidence, which puts you back to deciding on impressions.

Where does price enter?

After the tiering, not before. A price is only comparable between suppliers whose Tier 1 evidence shows they are quoting the same scope — same process ownership, same test coverage, same stackup intent. Comparing prices across different scopes is the most expensive mistake available at this stage, and it looks exactly like diligence.