A cosmetic surface class is the cheapest clause you will ever add to a sheet metal drawing and the most expensive one you will ever leave out. We build enclosures for EV chargers, battery cabinets, HMI terminals and telecom racks, and the arguments about appearance almost never start on the shop floor — they start at goods-in, when two people look at the same door and disagree about whether the mark on it matters. This article is about turning "it has to look good" into something an inspector can measure, a supplier can quote and a receiving bay can accept or reject without a phone call. We stay on the metal: how to zone a drawing, how to size a limit, why the same scratch passes on one finish and fails on another, and what each extra class actually costs in operations rather than in material.
Key takeaways
- There is no international standard that fixes cosmetic defect limits for sheet metal enclosures. Published supplier specifications commonly use a lettered class scheme (A through D or A through E). Those are OEM and supplier workmanship documents — a few are publicly downloadable, most are not — and none is an ISO, EN or ASTM standard. You have to write your own.
- A cosmetic class is defined by the viewer, not by the geometry: who sees the surface, how often, from how far, for how long, and under what light.
- Zone the drawing from the use position. A surface that faces a wall in service is not a Class A surface, however visible it looks on the 3D model.
- Limits have three axes: size, count and spacing. A limit with only one axis is unarguable in the wrong direction — a shop can pass a part full of small defects and still meet it.
- Visibility depends on the finish. A defect that disappears in a textured matte powder coat is obvious in a high-gloss or a brushed stainless skin, so the class and the finish have to be specified together.
- The cost of a higher class is labour and yield, not material: weld blending, local polishing, 100% visual inspection, individual packaging and a rework allowance.
- Limit samples end arguments. A signed physical sample photographed and stored with the part number is worth more than any sentence on a drawing.
A cosmetic surface class is a written acceptance grade applied to a defined area of a part. It states which appearance defects are tolerated on that area, how large they may be, how many, how far apart, and under what viewing conditions an inspector judges them.
Why "looks good" is not a specification
Every fabrication enquiry we receive that includes the phrase "cosmetically clean" or "no visible defects" is a future dispute. Those words mean whatever the reader wants them to mean on the day. A fabricator reading "no visible defects" has to price the part as though every visible face were Class A, because any looser reading is the one that gets argued at goods-in. You pay for that on the plinth, the rear door inner skin and the cable chamber — surfaces nobody has looked at since the unit was bolted down.
The opposite failure is worse. If the drawing says nothing, the shop applies its own default, and defaults differ. Some published supplier specifications state that all surfaces are Class C unless noted; others state that the default for inspection purposes is the second-lowest class. Two suppliers, same drawing, different outcome — and neither is wrong, because the drawing did not say.
So the first job is not picking numbers. It is deciding who gets to decide.
The three questions that actually define a cosmetic class
Forget the geometric description for a moment. Every published class scheme we have seen — and we have read a good number of them in customer specifications — works from three questions:
- Who sees it, and in what position? A customer-facing door skin on a payment terminal is seen by a member of the public at arm's length. The inside of the same door is seen by a technician with a torch, once a year.
- How close, for how long? An operator standing at an HMI panel all shift is a very different inspector from a passer-by walking past a telecom cabinet.
- What happens if it is marked? A scratch on a decorative cover is a return. A scratch on a sub-plate inside a control cabinet is nothing.
Answer those three and the class follows. Try to answer them from the 3D model and you will over-specify.
The five-class model and where it comes from
The lettered scheme used across the industry did not come from a standards body. It came from generations of internal workmanship documents at OEMs and contract manufacturers, and it has converged on a recognisable shape.
Class A through E, defined by viewer rather than geometry
| Class | Who sees it | Typical surfaces on an enclosure | What the limit usually means |
|---|---|---|---|
| A | Always seen, at close range, by the end user in normal use | Front door skin, HMI bezel, display surround, brand panel | Smallest limits, fewest defects, tightest spacing |
| B | Usually seen, at normal working distance, not the primary focus | Side panels, top surface of a low cabinet, visible hinge side | Larger limits, more defects permitted |
| C | Seldom seen — visible when a cover is removed or the unit is serviced | Interior walls, sub-plate faces, cable way covers | Coverage for protection; appearance secondary |
| D | Not visible in normal use; seen only during disassembly | Mounting flanges behind panels, internal bracketry | Functional protection only |
| E | Never visible in the assembled product | Interfaces buried in a sealed joint | No cosmetic criteria at all |
Note what the table does not say: it does not say "Class A is the front". It says Class A is what the end user looks at. On a floor-standing cabinet with a rear service door, the rear door can be Class A for a field technician and Class B for everyone else, and that is a legitimate split.
Some published specifications use four classes, some five. The labels are not standardised; the logic behind them is. What you are buying is the logic, not the letters.
Set your default class in writing
This single line prevents more disputes than any limit table. State, on the drawing or in the accompanying specification: "Surfaces not otherwise designated shall be Class C." Then mark the exceptions with a zone outline and a leader. Now the shop knows what the unmarked 90% has to be, and you have only drawn boundaries around what matters.
If you do not write the default, the shop will pick one, and you will find out which one it picked when the parts arrive.
How to zone a drawing: a repeatable method
Start from the use position, not the 3D model
Open the model in the orientation the product will actually occupy. For a wall-mounted HMI terminal that is mounted at eye level, the top surface is above the sight line and barely matters; for the same terminal mounted on a low pedestal, the top is the primary viewing surface. Same part, different zones.
Then walk through four positions:
- Normal use — the position a user occupies to operate the product.
- Purchase — how it looks on a pallet, in a box, on a showroom floor. Packaging damage shows up here.
- Service — what the technician sees with the door open.
- Never — everything else.
Surfaces that only appear in the last two rows are not Class A, whatever the render looks like.
Zones travel with the assembly, not the part
A common error is to zone individual sheet metal parts and then discover that two Class C parts assemble into a Class A face. Zone the visible surface of the assembly, then flow the requirement down to the parts that create it.
Practically: shade the assembly drawing, not the flat pattern. On the part drawing, put the zone boundary on the folded view, and add a note that the boundary is measured on the finished, assembled product. Otherwise a bracket that is perfectly compliant on its own will be rejected because the panel next to it is Class A — and that argument is unresolvable after the fact.
The defect taxonomy: name it before you can limit it
You cannot set a limit on "damage". You set limits on named defect types, and the names have to mean the same thing to the inspector and to the fabricator.
Substrate defects, coating defects and assembly defects
Grouping by origin tells you who owns the fix:
- Substrate defects originate in the metal or in forming: scratches from handling, dents from press-brake setup, roller marks from the mill, die marks that repeat on every part in the same place, burrs from cutting, and witness marks from forming tooling.
- Coating defects originate in pretreatment or application: runs and sags, orange peel, pinholes and outgassing craters, dry spray, colour or gloss variation between batches, thin coverage on sharp edges, and contamination inclusions.
- Assembly defects originate after finishing: fastener scuffs, hinge rub marks, scratch tracks from sliding a panel across a bench, fingerprints etched into a fresh coating, and mismatch at panel joints.
The reason the grouping matters commercially is that each family has a different owner and a different fix. A repeating die mark is a tooling problem and will affect every part in the run; a random handling scratch is a process discipline problem and affects a few. If your specification lumps them together, you will end up negotiating a blanket rework that fixes neither.
Table — defect types and what they actually look like
| Defect | How to recognise it | Usually caused by | Who owns it |
|---|---|---|---|
| Scratch | Linear removal or displacement of material; often catches a fingernail | Handling, nesting, bench contact, packaging | Fabricator / packer |
| Scuff or abrasion | Change in surface texture without material removal | Rubbing contact, sliding on a rack | Finishing line |
| Dent | Local depression, edges soft, no material removed | Impact during handling or assembly | Fabricator / assembler |
| Tooling or die mark | Depression or line that repeats in the same place on every part | Worn or damaged tooling, press setup | Fabricator (tooling) |
| Burr | Raised sharp edge at a cut or pierced hole | Cutting parameters, worn punch/die | Fabricator |
| Orange peel | Texture resembling citrus skin in the coating | Film build, cure profile, powder chemistry | Finisher |
| Run or sag | Localised thick coating with a downward edge | Over-application, part geometry | Finisher |
| Pinhole / outgassing | Small craters in the cured film | Gas escaping from substrate or trapped fluid during cure | Design or finisher |
| Colour or gloss shift | Visible difference against a reference panel | Batch variation, cure temperature, film thickness | Finisher |
| Weld discolouration | Straw-to-blue tint adjacent to a weld | Heat input and shielding during welding | Fabricator |
| Grinding witness | Local flat or swirl where a weld was blended | Blending method and grit sequence | Fabricator |
The last two are the ones that cause the most friction on enclosures, because both are inevitable products of the process and neither is a defect until the class says so.
Setting numeric limits that a shop can inspect
Size, count and spacing — the three axes
A usable limit statement has three numbers. Anything with one number is a loophole:
- Size — the maximum dimension of a single defect, usually expressed as width × length for a scratch or as a diameter for a spot.
- Count — how many defects of that size are permitted within the defined viewing area.
- Spacing — the minimum distance between two acceptable defects. This is the axis most specifications omit, and it is the reason a part can pass a written limit and still look terrible: thirty tiny scratches, each within size, spread across a door.
Add a fourth rule that costs nothing and closes the last gap: the total number of mixed defects shall not exceed the count limit of the largest defect type present. Published supplier specifications that use this wording are worth copying verbatim, because it prevents the "three of each" reading.
Viewing conditions: distance, time, light and angle
A limit without viewing conditions is meaningless, because every defect disappears if you step far enough back.
Published examples from supplier workmanship documents typically define conditions along these lines:
| Variable | Tighter classes | Looser classes | Notes |
|---|---|---|---|
| Viewing distance | Around 450 mm | Around 450–600 mm | Arm's length is the common reference for smaller parts |
| Viewing time | Around 10 s per surface (the most commonly published value) | Around 5–10 s per surface | Longer inspection finds more, not better, parts |
| Lighting | Uniform, non-directional, cool white; published minimums commonly fall in the 80–150 foot-candle band | Same | Directional light turns texture into apparent defects |
| Viewing angle | Surface tilted through a range (often 0–45°) | Normal to the surface | Tilting reveals dents and gloss shift |
| Magnification | Not used | Not used | State this explicitly; it is the most common escalation |
| Assessor vision | Normal or corrected-to-normal | Same | Rarely stated, worth stating |
Treat these as illustrations of what a complete clause looks like, not as values to copy. Adopt your own, put them in the specification, and make sure the receiving bay uses the same ones. If colour or gloss matching matters, ASTM D1729, the Standard Practice for Visual Appraisal of Colors and Color Differences of Diffusely-Illuminated Opaque Materials, is a useful reference for how to set up the appraisal conditions. It does not define defect limits, which is exactly the gap you have to close yourself.
For surface texture symbols on the drawing itself — as opposed to appearance defect limits, which sit alongside your dimensional tolerances — the notation is governed by ISO 21920-1:2021, which replaced the withdrawn ISO 1302:2002, or by ASME Y14.36-2018 in the US system. Those tell you how to draw the symbol; they do not tell you what appearance defects are acceptable. Conflating the two is the single most common error we see in customer specifications.
Table — example acceptance matrix (illustrative only)
The numbers below are a starting draft for a carbon steel or stainless steel enclosure with a matte powder coat. They closely follow the shape of published supplier workmanship tables and are not a standard; adjust them with your supplier before you issue. For how substrate choice interacts with what you are able to achieve, see our comparison of stainless steel and carbon steel enclosures.
| Defect | Class A | Class B | Class C | Class D |
|---|---|---|---|---|
| Crack | None | None | Short, limited count | Limited count |
| Dent or pit | None | Small, isolated, no exposed base metal | Medium, no exposed base metal | No limit if base metal protected |
| Scratch | ≤0.1 mm wide × ≤3 mm long, max 1 | ≤0.1 mm × ≤20 mm, max 2, ≥30 mm apart | ≤0.2 mm × ≤75 mm, max 2 | No limit if base metal protected |
| Discolouration | None | ≤0.5 mm diameter, max 2 | ≤2 mm diameter, max 2, no colour shift against reference | No limit if corrosion protected |
| Weld discolouration | Blended and matched | Visible tint acceptable at defined distance | Acceptable | Acceptable |
| Coating run or sag | None | None | Minor, outside sealing faces | Acceptable if function unaffected |
| Exposed base metal | None anywhere | None | None | None |
The last row is the one rule that should never relax by class. Exposed base metal is a corrosion requirement, not a cosmetic one — see our guide to sheet metal surface finishing for what coverage actually protects.
Visibility is a function of finish, not just defect size
The same 0.2 mm scratch is invisible on a wrinkle-textured powder coat and glaring on a polished stainless skin. Specifying a class without specifying the finish alongside it is specifying half a requirement.
Table — finish × defect visibility
| Finish | What it hides | What it reveals | Practical consequence |
|---|---|---|---|
| Texture / wrinkle powder | Dents, grinding witness, minor flatness variation | Coating thickness variation, colour shift | Best choice when substrate quality cannot be Class A |
| Matte smooth powder | Small dents, minor orange peel | Scratches, handling marks, edge thinness | The default for most industrial enclosures |
| Semi-gloss / gloss powder | Almost nothing | Dents, orange peel, flatness, every handling mark | Treat any gloss surface as effectively one class tighter |
| Brushed stainless (No.4) | Small pits, minor tint | Anything across the grain, inconsistent grain direction | Grain direction must be specified, not implied |
| 2B mill finish stainless | Scratches | Roller marks, flatness, weld tint | Avoid on visible skins; see our stainless steel surface finish guide |
| Anodised aluminium | Light scuffs | Any substrate scratch, dye lot variation, rack marks | Bath lot consistency matters more than defect size — see our aluminium anodizing guide |
| Zinc-plated / galvanised | Minor handling marks | Drain and vent marks, rack contact points, ash | Geometry drives appearance; see our powder coating hang points, vent holes and masking guide |
The practical rule: if your design needs a high-gloss or polished finish, budget for the substrate quality of one class higher. Gloss does not fix substrate; it publishes it.
Grain direction and why it beats any numeric limit
On brushed or polished stainless and on aluminium with a directional finish, the single most effective cosmetic clause is not a number at all. It is a grain direction arrow plus one sentence: "Grain direction as shown; all blending and polishing to follow grain."
An inspector can check that in two seconds and cannot argue about it. A 3 mm limit on a brushed panel can be argued for an hour. Where the appearance of a directional finish matters, spend your specification effort on grain, not on millimetres.
This applies to formed features too. A bend across the grain behaves differently from a bend with it in terms of surface stretch, and the witness line left by the press-brake die sits differently relative to the grain. Our sheet metal bending guide covers the mechanics.
Where cosmetic class meets the process
Weld blend and polish levels
On a welded enclosure, the appearance of the visible skin is decided at the weld, long before coating. Three levels are worth distinguishing in writing:
- As-welded — weld bead and heat tint remain. Acceptable on internal structure, never on a Class A skin.
- Blended — bead removed flush with the parent metal; a local change in texture remains visible after coating.
- Polished / matched — blended then brought up through grit stages to match the surrounding texture and, on stainless, passivated afterwards.
In our shop, each step roughly doubles the labour at that joint; your supplier's ratio will differ with joint access and volume. Specify the level per zone, not globally: an enclosure that is polished on the front face and as-welded inside will look right and cost sanely. One that is polished everywhere will cost a fortune and nobody will ever know.
If the part is stainless, blending has to be followed by the chemical step — see our guide to stainless steel passivation and weld cleanup. The tint itself comes from heat input and shielding during welding. If it is carbon steel going to powder coat, note that coating does not fill a dent and does not hide a grind witness under gloss.
Weld sequence matters too, because distortion on a Class A skin is a cosmetic defect as much as a fit problem — see our weld distortion control guide.
Handling, racking and packaging
Most cosmetic rejects are not fabrication defects. They are handling defects: parts touching on a rack, sliding across a bench, stacked without interleaving, or packed without edge protection. Once you have specified a Class A zone, the process has to follow it into the packing station.
Ask for these four things and you remove most of the risk — the fourth is the one that decides whether the incoming inspection finds anything at all:
- Rack contact points on non-cosmetic surfaces only, agreed before the first article, and consistent between the sample and production.
- Interleaving or sleeving on Class A faces, in the process, not just in the carton.
- Individual protection for Class A assemblies, with corner and edge protection rather than bulk wrap.
- A packing method that survives the actual route — see our export packaging and crating guide for what a multi-leg shipment does to a finished face.
Table — operations added by each class
| Operation | Class A | Class B | Class C / D |
|---|---|---|---|
| Weld finish | Blend and polish, matched to grain | Blend flush | As-welded, remove spatter |
| Local rectification | Permitted with defined method and touch-up limits | Permitted | Not required |
| Inspection coverage | 100% visual on the zone | Defined sampling | Routine |
| Viewing conditions applied | Strictly, with limit samples | Applied | Not applied |
| Handling discipline | Gloves, interleaving, dedicated racks | Standard | Standard |
| Packaging | Individual protection, edge and face protection | Standard with separation | Standard |
| Rework allowance | Built into the quote | Built into the quote | Not applicable |
What it costs and where the money goes
The premium is in labour and yield, not material
A higher cosmetic class does not add steel. It adds touch time: blending, polishing, extra handling steps, individual packing, 100% inspection, and the rework. The rectification allowance is usually the biggest line, because at Class A the first-pass yield on a welded door skin drops sharply.
When you ask a supplier to quote Class A and Class B versions of the same door, the difference will not appear as a clean percentage — the same is true of every enclosure cost driver, but appearance is the one buyers most often leave unquoted. It will appear as line items: blending labour per metre of weld, polishing consumables, inspection time per unit, upgraded packing, and a scrap or rework allowance derived from the expected yield at that class. Ask for both quotes with the line items visible, and the decision becomes a commercial one you can make with facts.
When to downgrade a zone and save real money
Walk the assembly and look for these five, because each is usually mis-zoned upward:
- The top of a tall cabinet. Above the sight line for anyone standing at it.
- The rear face of a wall-mounted or rear-serviced unit. Seen during installation, then never again.
- Surfaces behind a fixed shroud or sunshade. Our outdoor sunshade and hood design guide shows how much area these hide.
- Lower side panels below about knee height on a floor-standing unit in a plant environment, where they will be scuffed in service anyway.
- Any face inside a sealed joint that cannot be seen without destroying the assembly.
Downgrading each of those from A to B typically removes more cost than any material or process change available at that point in the design.
Resolving disputes: limit samples and golden samples
Every cosmetic argument that reaches a phone call is a failure to have made a sample.
How to run a limit-sample programme
- Make the sample at the start of production, not at the dispute. Use real production parts, from a real run, on the real finish.
- Make two: the best acceptable part and the worst acceptable part. "Golden sample" alone is a trap — it sets the bar at the best part anyone will ever see.
- Mark the defects on the worst-acceptable sample with a fine ring and a number, and record what each one is and why it passes.
- Sign and date both, photograph them under the specified lighting, and store the record with the part number.
- State the review rule: samples are re-made at every finish change, every material change, and every twelve months otherwise.
- Hold one at each side. If the receiving bay and the fabricator each hold a matched pair, disputes end in minutes.
Where an enclosure programme runs for years, link the sample set to the revision state. When the drawing moves, the samples have to move with it — the same discipline we describe in our guide to drawing revisions and engineering change control.
A spec you can paste into a drawing
Adapt the following to your drawing standard. Every clause exists to remove a decision from the moment of inspection.
- "Cosmetic surface classes per this drawing. Surfaces not designated shall be Class C."
- "Class A: door outer skin, display surround and brand panel as shaded. Class B: side panels and top. Class C: all interior surfaces."
- "Inspection conditions: uniform non-directional lighting [specify level], viewing distance [specify], viewing time [specify] per surface, part in the use position, no magnification."
- "Defect limits per class: size, count and minimum spacing as tabulated. Total mixed defects shall not exceed the count limit of the largest defect type present."
- "Exposed base metal is not acceptable on any class." — this is a corrosion protection requirement wearing a cosmetic badge.
- "Weld finish: Class A blended and polished to match adjacent texture; Class B blended flush; Class C as-welded, spatter removed."
- "Grain direction as indicated. All blending and polishing to follow grain."
- "Rack and handling contact points to be on non-cosmetic surfaces. Class A faces to be interleaved and individually protected."
- "Limit samples: two per class, agreed before first article, reviewed at every finish or material change."
- "Colour and gloss to be appraised against the agreed reference panel under the specified lighting."
Then quote it with the rest of the commercial package — our sheet metal RFQ checklist lists what else has to travel with the drawing, and our first article inspection guide covers how to make the first batch the reference for every batch after it.
If you would rather have us mark the zones for you, send the model and tell us where the product lives and who stands in front of it. We will shade the assembly, propose the classes, quote both options with the line items visible, and tell you which zones we would downgrade.
Frequently Asked Questions
Is there an ISO or EN standard that defines cosmetic surface classes for sheet metal?
No public international standard fixes cosmetic defect limits for sheet metal enclosures. The lettered class schemes you see in circulation originate from individual OEM and contract-manufacturer workmanship documents, and the numbers inside them are company property. Standards such as ISO 21920-1:2021 (which replaced the withdrawn ISO 1302:2002) and ASME Y14.36-2018 govern how surface texture is indicated on a drawing; they do not define acceptable appearance defects. Write your own classes and limits into the drawing.
What is the difference between Class A, B and C surfaces on an enclosure?
Class A surfaces are always seen by the end user at close range in normal use — a door skin, a display surround, a brand panel. Class B surfaces are usually seen but are not the primary focus, such as side panels. Class C surfaces are only seen during service or disassembly, such as interior walls. The class is set by who sees the surface and for how long, not by where the surface happens to be on the model.
What should a cosmetic defect limit actually specify?
Three axes at minimum: the maximum size of a single defect, the permitted count within the defined viewing area, and the minimum spacing between two acceptable defects. Add a rule that the total number of mixed defects cannot exceed the count limit of the largest type present, or a part covered in small defects will still pass.
Do I need viewing conditions in a cosmetic specification?
Yes, and they are the clause most often missing. Without a stated viewing distance, viewing time, lighting and angle, any defect can be argued away by stepping back. Published supplier examples commonly use arm's-length distances, a few seconds per surface, uniform non-directional cool-white lighting and no magnification. Adopt your own values and make sure the receiving bay uses the same ones.
Why does the same scratch pass on one enclosure and fail on another?
Because visibility depends on the finish. A textured or wrinkle powder coat hides small dents and grinding witness marks, while a gloss or polished surface reveals flatness variation and every handling mark. A high-gloss or polished skin effectively behaves one class tighter than a matte one on the same substrate, so specify the class and the finish together.
Does a higher cosmetic class increase material cost?
Not materially. The premium is labour and yield: weld blending and polishing, 100% visual inspection, interleaving and individual packaging, and the rework or scrap allowance from the lower yield at that class. Ask for the Class A and Class B versions quoted side by side with the line items visible, and the trade-off becomes a commercial decision you can make with facts.
How do limit samples prevent cosmetic disputes?
Make two real production parts before volume production: the best acceptable part and the worst acceptable part. Ring and number the defects on the worst-acceptable sample, record why each passes, sign and photograph both under the specified lighting, and hold a matched pair on each side. A dispute then resolves in minutes against a physical reference instead of by negotiation over a sentence on a drawing.
Can a cosmetic change affect an existing first article report?
Yes. A change to finish, material, grain direction or weld finish changes the appearance result and can invalidate the approved reference. Treat it as an engineering change, re-run the relevant part of the first article, and re-issue the limit samples against the new revision.