Surface Finish Specification and Verification on Custom Metal Parts

Unclear surface finish callouts cost time and money before any metal is cut.

· 9 min read · Updated
Cover illustration for “Surface Finish Specification and Verification on Custom Metal Parts”
Quality Assurance · October 2, 2026 · 9 min read · 1,977 words

Most surface finish rejections and rework cycles start on paper, not on the factory floor. A callout that just says "black anodize" specifies nothing. It leaves open which anodize type, what thickness, which surfaces count, whether the toleranced dimensions apply before or after finishing, and what corrosion performance the part actually needs to survive. Every one of those open questions gets answered by somebody at the supplier, and they'll answer it the cheapest way the drawing allows. The supplier built exactly to what the drawing said, so there's no real ground to fight the rejection on. When a drawing reads "surface finish: TBD" or names only a color, the factory picks the process, and what they happen to have running that week may have nothing to do with the part's environment, its tolerance stack, or what it needs to do in service. The fix is a callout that closes every gap the supplier would otherwise have to guess at, and that work happens at the drawing, before a single part gets cut.

What a surface finish does for a metal part

A surface finish does a job. Picking the wrong process for that job is a specification mistake no matter how well the supplier runs it. A finish on a machined part does one or more of five things: it protects against corrosion, it adds surface hardness and wear life, it changes electrical behavior, it controls appearance, or it manages friction at a sliding interface. Processes that are great at one of these are often wrong for another. Hard anodize makes an excellent wear surface and an excellent electrical insulator. It is the wrong choice for a part that needs to ground through its mounting face. Zinc plate protects steel well but is too soft to carry a bearing load. None of this is about grading a part as high quality or low quality. Name the one function the finish actually needs to deliver, because that naming is most of what specifying a finish requires. One more distinction matters here: machined surface texture and applied finish are two separate requirements. Treating them as a single callout is a common and avoidable source of ambiguity.

How finishing processes affect a part's dimensions and performance

Each finishing process changes a part's dimensions in its own way and fails in its own way when someone picks it for the wrong job. Anodizing, including hard anodize, grows an oxide layer out of the base metal itself. Roughly half the coating thickness comes out of the original surface and half builds up above it. On a close-fitting bore, that's a real tolerance shift, not a cosmetic detail, and the fix is to machine the feature undersize before finishing so the growth brings it to nominal. A complete anodize callout needs three things on the drawing: the coating class or type, the thickness, and the seal.

Electroplating adds metal using applied current, and that current doesn't spread evenly. Outside corners plate thicker, while deep bores plate thin or sometimes not. Where a bore has to be protected, electroless nickel is the steadier choice because it deposits through a chemical reaction rather than current flow, so the coverage stays even front to back.

Conversion coatings like chromate and phosphate add almost no measurable thickness. The conductive versions keep a surface electrically usable, which makes them the right call when a part has to be protected and still grounded at the same time.

Passivation of stainless steel is a cleaning and chemical treatment, not a coating, and it adds no measurable thickness either. Corrosion performance here is a test result, not an adjective. "Corrosion resistant" on a drawing isn't a specification, and it can't support a rejection on its own.

Powder coating and paint add real, measurable thickness, so masking has to be called out as its own line item rather than assumed. Threads, bearing bores, sealing faces, and electrical contacts all need to be named on the drawing if they're supposed to stay bare.

Subtractive and refining steps, electropolishing, abrasive blasting, buff polishing, get used when the goal is lower surface roughness, burr removal, or an even texture across the part. Most additive finishes need one of these as a prep step first, so the sequence on the drawing matters as much as the finish itself.

For alloy steel tubes in hydraulic, pneumatic, or fatigue-sensitive service, surface finish and post-treatment aren't optional add-ons. They need to be called out explicitly, right alongside heat treatment and NDT requirements, because the surface condition drives fatigue life in those applications just as much as the material spec does.

Where Ra fails as a roughness specification

Ra works fine for general surfaces, but it fails on sealing faces and bearing journals, where one deep scratch that Ra simply averages away is enough to cause a leak or start a premature wear failure. Ra, the arithmetic average roughness, gives a single number that represents overall roughness across the surface. Because it's an average, it smooths right over an occasional deep gouge or an isolated high spot. Rz, the maximum height of profile, instead captures the largest peak-to-valley distance within each sampling length, which makes it far more sensitive to the defects that actually cause functional problems. Ra and Rz come from the same surface profile trace, but they're calculated in fundamentally different ways. Ra averages every deviation from the mean line. Rz averages the peak-to-valley heights across the sampling lengths. On a sealing surface, calling out Ra alone leaves a gap in the spec that calling out Rz would close. Someone will point out that Rz is harder to measure and harder to agree on between engineer and supplier, and that's true. It still isn't a reason to leave it off the drawing, because the alternative is a callout that waves through a defective part.

Non-obvious surface finish requirements on sealing interfaces and gasket faces

Sealing surfaces and gasket faces need a roughness range, not just a ceiling, because a surface that's too smooth fails by a completely different mechanism than one that's too rough. Extremely smooth surfaces don't give a gasket enough grip to resist lateral shear from internal pressure or thermal expansion and contraction. The gasket can slip sideways under load, and once it slips, the seal is broken, even though nothing about the surface looks defective by eye. A controlled amount of roughness gives the gasket material something to key into as it compresses, so the texture is doing real structural work, not just sitting there as a side effect of machining. That means the callout needs both a maximum and a minimum Ra or Rz, instead of a vague instruction to make the surface "smooth enough. The same logic carries over to any interface where a rubber or soft-metal seal has to grip a machined face under dynamic load. Most DFM checklists warn against surfaces that are too rough, but standard practice rarely addresses the risk of over-polishing a sealing face. That gap in common guidance is why it belongs on the drawing explicitly rather than left to habit.

Coating thickness and dimensional tolerances at drawing stage

Coating thickness is never a rounding error on a precision fit. A drawing that sets a tight tolerance without saying whether it applies before or after finishing has written a contradiction into the part. Machined surface texture and applied finish are separate requirements, and the safest habit is to state, for every toleranced dimension, whether it's measured before or after finishing. Take anodizing: if roughly half the coating thickness grows into the surface, a bore machined to nominal and then anodized comes out undersized by that amount. The fix is to machine the pre-finish feature undersize so the finished part lands on nominal after the coating grows. Electroplating raises a different problem. Because current density isn't even, a deep bore can plate to a different thickness than an outside face on the same part, so the as-finished dimension on that bore can't be predicted from the nominal coating spec alone without knowing the part's geometry. The reliable approach: state the dimension as it should measure after finishing, give the coating thickness with its own tolerance, and machine the pre-finish feature to land on that as-finished nominal once the coating is applied. A complete finish callout on a fabricated part needs to name the finish type, the color or gloss, the protected areas, and the dimensional impact, all as required elements rather than optional detail. And masking has to be treated as its own line item, not an assumption. Threads, bearing bores, sealing faces, and electrical contacts all need to be named if they're supposed to stay uncoated, because leaving that unspecified hands the supplier discretion over which surfaces get coated and which don't.

The seven elements a complete finish callout must contain

A finish callout missing any of these seven elements has a gap, and the supplier will fill that gap with whatever interpretation costs the least. The first element is process and class or type. This is what separates a decorative coating from a wear-rated one, Type II anodize against Type III hard anodize being the clearest example, and the word "anodize" by itself doesn't tell anyone which one is meant. The second is thickness with a tolerance, since thickness sets both the corrosion life of the part and how much it grows dimensionally, and it needs to be given as a range rather than a single nominal number because every coating process carries some natural variation. The third is a clear statement of which surfaces get finished and which don't, naming them directly rather than relying on "all over" or leaving the question blank. The fourth is a corrosion requirement stated as a testable number: a test method and a duration, because "corrosion resistant" is an adjective and can't be the basis of a rejection on its own. The fifth is color and gloss, given in terms that can actually be measured or matched against a reference standard rather than described in casual language. The sixth is the seal or post-treatment step, since for anodizing in particular, whether the pores are sealed and by what method determines the part's long-term corrosion performance. The seventh is the dimensional basis, a plain statement of whether toleranced dimensions apply before or after finishing, and this single missing item is behind more fit-up failures after coating than any other cause on this list. When an engineer sends out an RFQ for finished custom parts, the quote request itself should spell out which of these steps the factory controls, the surface preparation method, the finish specification, and the inspection criteria, because leaving any of them open hands the factory discretion the engineer almost certainly doesn't want them to have.

The over-specification risk: when tighter finish callouts add cost and inspection burden without reducing functional risk

Every surface finish requirement that doesn't need to be there adds machining time, raises the odds of scrapping a part, and drives up inspection cost. The goal is complete specification where it actually matters, not tight specification everywhere. A tight Ra callout or a profile requirement placed on a surface that does no functional work in the assembly adds cost to check without lowering any real risk, and every GD&T callout on a drawing should trace back to something the assembly actually needs from that surface. DFMA studies on manufacturing cost consistently show that removing one unneeded feature cuts part cost more reliably than haggling over raw material prices, and an unnecessary finish callout is exactly that kind of unneeded feature. The engineer's job is to know which surfaces are doing real work, sealing, bearing, grounding, wearing, and put the seven-element callout on those, while leaving the rest of the part to a general finish note that doesn't pretend to control dimensions nobody is relying on.

Sources

  1. Surface Roughness Average (Ra): Chart, Formula, and Conversion Calculator
  2. Surface Roughness: A Comprehensive Guide for Engineers
  3. A computational framework for predicting the effect of surface roughness in fatigue
  4. Guide to Surface Finishes and Tolerances
  5. How Surface Height Is Quantified — From Ra to Rz
  6. Surface Roughness Explained: Ra vs Rz vs Rq - Kemet
  7. Flange sealing surface finish requirements
  8. Flange Surface Finish Guide: Ra, Roughness, and Gasket Sealing - Sunhy

More in Quality Assurance