How do I judge surface finish capability from a supplier’s samples?

Introduction

A surface finish requirement looks like the simplest line on a drawing: Ra 0.8, maybe Ra 1.6, occasionally a note about the direction of the lay. In practice that one line is where a surprising number of CNC sourcing decisions go wrong, because the sample that convinced you and the batch that arrives are produced by two different processes. A supplier’s showpiece is evidence of skill, not evidence of control, and at production volume control is the only thing that protects your sealing faces, bearing bores and coating adhesion. The purpose of this article is to give procurement teams a way to read a finish claim as a process statement, so that a pleasing coupon stops being the deciding factor.

At Ruicheng Industrial Design Co., Ltd., where CNC milling, reverse engineering, finishing and CMM inspection sit inside one ISO 9001 workflow, the question we are most often asked is not whether a finish can be achieved but whether it can be held across a batch and repeated six months later. That distinction shapes everything below: which parameters actually describe the surface, how to test a claim against real production conditions, and what a capable supplier commits to in writing before the order. The same logic applies whether you are buying a new housing or reproducing a discontinued one from a worn sample, because in both cases the finish that matters is the one on the part you have not received yet.

Which surface parameters actually separate a capable shop from a persuasive sample?

A supplier’s showpiece sample is usually the single best part they have ever made, which is precisely why it is a poor predictor of your production finish. A mirror-like coupon can be hand-lapped for an hour by the most patient operator in the building, while your housing needs the same result across four hundred pieces, three setups and two anodising batches. Judge the process that produced the sample, not the sample itself, because a finish you can reproduce is a capability and a finish you can only demonstrate is a performance. The distinction matters most on functional surfaces: sealing faces, bearing bores and optical mounts behave according to their roughness profile, not according to how they photograph under a bench light.

Ra is an average, not a guarantee: A single Ra number collapses a whole profile into one value, so two surfaces with identical Ra can behave completely differently in sealing, wear or coating adhesion because one is plateaued and the other is sharply peaked; ask for Rz or Rpk alongside Ra, and ask which direction the lay runs relative to the sealing path.

Ask what the sample was made on: A finish achieved on a finishing mill with a fresh wiper insert and a light final pass is a production process, while the same number reached by manual polishing, abrasive flap wheels or a post-process tumble is a labour operation that will vary with the operator, the shift and the consumable batch.

Separate machined finish from post-process finish: Bead blasting, anodising, plating and passivation all change the measured roughness, and a supplier quoting Ra 0.4 before anodising may be describing a surface that ends at Ra 1.2 after the coating you actually specified; the finish requirement belongs on the finished part, not on an intermediate state.

Require the measurement record: A profilometer trace with the cutoff length, evaluation length, stylus radius and measurement direction is evidence, whereas a number typed into an email is an assertion; a shop that measures routinely can produce the trace in minutes and will usually show you the setup that generated it.

A buyer who wants a defensible answer should ask the supplier to walk through a documented CMM inspection workflow and then explain how their reverse-engineering and replication process holds surface specifications when the original drawing is incomplete. The shops that can narrate both are the ones whose finish claim survives contact with a production batch, because they treat roughness as a controlled output of a defined process rather than as an aesthetic achieved at the end. This is also where a general capability overview helps you check whether the finishing equipment is actually in house or subcontracted, since a subcontracted finish adds a handoff where the traceability usually breaks.

🔧 A beautiful sample proves someone on the floor is skilled; a repeatable trace proves the shop is capable.

How do I check that the claimed finish survives real production conditions?

The honest test of a surface finish claim is not a first article but a batch: the twentieth part, the second tool insert, the Monday morning shift and the lot that ran after the coolant was changed. Capability is the ability to hold a specification when the easy variables stop being favourable, and the only way to see that is to ask for evidence sampled across time rather than a single golden part. Tool wear alone can shift Ra by a factor of two on a finishing pass, and a shop without a defined insert-change interval will drift through your tolerance band long before anyone notices.

Request parts from different points in a run: Ask for three samples taken at the start, middle and end of a production lot rather than three from the same tray, then measure them yourself or send them to one independent lab; the spread between them is a far better capability signal than the absolute value of any one of them.

Look for a defined tool-life rule: A mature supplier changes a finishing insert by a counted part interval or a measured wear land, not when the operator hears a change in cutting sound, and they can tell you the interval they used for your part and why.

Check the coolant and burr policy: Coolant concentration, filtration and the deburring method all leave their own signature on the surface, and a shop that controls them has a written routine that covers the whole range of parts it runs rather than a habit that works for the easy ones.

Tie the finish to a functional test: Where the surface matters, specify the outcome as well as the number: a leak rate for a sealing face, a torque scatter for a mating thread, or a coating adhesion result for a prepared substrate, so the requirement cannot be met cosmetically while failing in service.

The practical way to structure this is to put the evidence requirement into the purchase order alongside the dimension plan, referencing how the supplier verifies capability through first-article reports and how tight-tolerance work is guaranteed on similar programmes. Ask for a short control plan that names the finishing operation, the insert change rule, the measurement device and the reaction plan when a reading drifts. A supplier who already has that document will send it without fuss; one who has to write it for you has just told you the finish is being managed informally.

✅ Ask for three samples from three points in the run; the spread between them is the real specification.

What does a finish-capable supplier put in writing before the order?

Most surface finish disputes are not measurement disputes; they are scope disputes. The drawing says Ra 0.8, the supplier assumes that applies to one flat face, and the buyer assumes it applies to every machined surface including the bore and the stepped shoulder. Write down where the requirement applies, in which state, measured how, and by whom, and nine out of ten finish arguments never happen. The same clarity is worth having about who owns the cost of a re-finish when a coating step changes the reading.

Define the surfaces and the state: Call out each functional surface on the drawing with its own requirement, and state whether the value applies before or after heat treatment, anodising or plating, because a requirement that silently moves with the process step is a requirement nobody can consistently meet.

Specify the measurement method: Name the instrument class, the cutoff, the number of traces and the direction of measurement relative to the lay, since a stylus trace across the tool marks and one along them can differ enough to turn an acceptance into a rejection.

Agree the acceptance and reaction plan: State the sampling frequency, the allowable drift before a tool change, and what happens to parts made between the last good check and the first bad one, because containment is where an informal process becomes expensive.

Put finishing in the same workflow: Prefer a supplier who machines, finishes and inspects under one quality system, so a failed reading triggers a program correction rather than a debate between two vendors about whose step caused it.

Before awarding, review how the shop handles reverse-engineering and replication when the original surface specification is missing, and how it evaluates capability on discontinued parts where the only reference is a worn sample. Those two cases are the hardest version of the problem, and a supplier who can reconstruct a defensible finish requirement from an aged component is unlikely to be vague about a drawn one. Our service overview sets out where milling, finishing and inspection sit in one workflow.

⚠️ A finish requirement without a measurement method is a preference, not a specification.

Surface finish claim maturity across four supplier types

The right column depends on what the surface has to do. For a cosmetic panel the second column may be perfectly adequate, but for a sealing face, a bearing bore or a coated optical mount the control plan is worth more than the sample, because it is the only thing that describes the parts you have not received yet.

The right column for a given programme depends on which of these rows you treat as non-negotiable, and a short technical review usually settles it faster than another round of pricing. For production parts where consistency matters more than the first-batch price, the controlled column almost always wins on total cost of ownership. If you want a second opinion on where your part sits, contact us and we will walk through the trade-offs with your drawing in front of us.

A practical qualification sequence for surface finish claims

A workable qualification is built in four passes, and collapsing them into one email is the usual reason a promising sample turns into a disappointing first batch. Fix the requirement, interrogate the process, sample the run, and only then compare prices.

1. Write the requirement in the finished state: Name each functional surface, its roughness parameters, the coating state it applies to and the measurement method, so every supplier is bidding the same specification rather than a convenient interpretation of it.
2. Ask how the sample was actually made: Require the machine, the finishing pass, the insert or abrasive used and whether any manual polishing was involved, because a hand-finished sample cannot be priced as a machined finish.
3. Sample across a real production run: Request three parts from three points in a lot and have them measured independently, then compare the spread as well as the absolute values.
4. Agree the control and reaction plan: Put the insert-change rule, the sampling frequency and the containment rule into the order, so a drift becomes a documented correction rather than a negotiation.

One detail that repays attention early is how a supplier handles the boundary between machining and finishing. On a housing with a sealing land and a separately blasted cosmetic face, the same nominal roughness can be produced by two entirely different routes, and only one of them is stable when the batch grows. A finishing pass with a wiper insert, a controlled step-over and a documented tool-change interval produces a surface whose texture is repeatable in direction and amplitude, while a manual blend after machining produces a surface whose appearance is excellent and whose profile is essentially random. The distinction is invisible in a photograph and obvious in a profilometer trace, which is why the trace belongs in the quotation package rather than in a later quality dispute. Buyers who specify the lay direction alongside Ra usually find that half of the candidates quietly drop out, because they were planning to polish rather than to machine.

It is also worth separating what you need to know at RFQ stage from what you need to verify at first article. At RFQ stage, three questions do most of the work: which process produces the finish, in which state is the value specified, and what measurement evidence will accompany the batch. Those answers let you compare quotes on a like-for-like basis and expose the supplier who has priced a hand-finished sample as though it were a machined output. At first article, the questions become narrower and more evidential: is the trace consistent with the quoted method, do parts from three points in the run agree, and does the agreed coating step still leave the surface inside the band. Keeping the two stages separate prevents a common failure where a buyer negotiates hard on a number at RFQ, only to discover at inspection that the number described an intermediate state nobody intended to deliver.

A final practical point concerns what happens when a reading drifts. Even a well-controlled finishing process will eventually approach the edge of a roughness band as an insert wears, and the cost of that drift depends almost entirely on whether the supplier has a rule for it. A shop with a counted change interval and a sampling plan catches the trend while every part in the batch is still acceptable, then changes the insert and documents the correction. A shop without one tends to notice at final inspection, which turns a ten-minute tool change into a containment exercise covering several hours of production. Asking to see that rule before the order is a small request that reveals a great deal about how the finish is actually managed, and it is usually the cheapest insurance available against a surface-related reject.

Frequently asked questions from procurement teams

Is Ra 0.8 always achievable on milled aluminium?

On a flat face with a fresh wiper insert and a controlled final pass, yes, and it is routine. On a deep pocket floor, a thin wall or a surface that will later be anodised, the same number can be difficult or irrelevant. Ask for the parameter in the state the part will actually be used, and ask which surfaces it covers, because a blanket Ra note on a drawing is a common source of both over-quoting and rejected lots.

Why did the measured finish change after anodising?

Anodising builds an oxide layer that both adds thickness and roughens the underlying profile, and the effect depends on the alloy, the pre-treatment etch and the coating thickness. A surface measured at Ra 0.4 as machined can read noticeably higher afterwards. Specify the value on the finished part, or state both values explicitly, and keep the pre-treatment step in the same supplier’s control plan.

Should I supply my own samples for the supplier to match?

Yes, if they are genuinely representative. Include parts from more than one production lot and, if possible, one part that is at the edge of acceptability, so the supplier understands the band rather than a single target. Photographs are useful for context but cannot substitute for a traceable measured sample, and a worn or polished used part should be identified as such rather than presented as a nominal reference.

What inspection evidence should come with the first article?

A profilometer report with the cutoff and evaluation length, the measurement direction relative to the lay, and the specific surfaces checked, alongside the dimensional CMM report. Ask for the trace itself rather than a pass or fail summary. A supplier who measures routinely will also be able to show how the reading sits against the tool-change interval used for that batch.

Can you help reconstruct a finish requirement from an old sample?

Yes. Where a discontinued part arrives without a usable drawing, we measure representative surfaces, separate wear from the original manufacturing texture, and propose a defensible requirement tied to the part’s function. Ruicheng Industrial Design Co., Ltd. keeps measurement, milling and finishing in one workflow, which makes that reconstruction considerably faster than sending the sample between three vendors.

Conclusion

A surface finish claim is only as good as the process behind it. The sample tells you what a supplier can do once; the control plan tells you what they will do four hundred times, and the second is the only thing your assembly line ever sees. Write the requirement in the finished state, ask how the sample was made, sample across a real run, and agree the reaction plan before the order rather than after the first reject. Do that, and the difference between a persuasive coupon and a capable shop becomes obvious long before a batch is at risk.

For expert assistance in implementing surface finish qualification for CNC-machined parts for your production needs, visit our resource center or contact us. Let’s help you scale up your manufacturing with precision and efficiency!