Elsewhere on this site we argued that, inside a specification, a performance class is a sentence rather than a number — because a class printed without the standard, the edition, the tested size and the operating type cannot be judged pass or fail, and a requirement that cannot be failed is not a requirement. This article starts where that one ended. Assume the sentence is complete. You are holding a designation that reads something like Class AW – PG40 – Size tested 1500 x 2500 mm – Casement. Now ask a different question: how much does that sentence actually tell you about the window in front of you?
Less than it looks like. A performance class is not a measurement. It is a floor. It is the rung a specimen reached on a ladder somebody else drew, and by construction it is the rung below the result — not the result itself. Almost everything a buyer wants to know lives in the distance between those two things, and almost no submittal package records it.
The clearest evidence is not in a marketing brochure; it is in the wording of the test standard itself. In GB/T 7106, the wind-resistance class is assigned at the pressure step immediately before the step at which dysfunction or damage appeared. The water-tightness class works the same way: the value of the specimen is the pressure step before the step at which severe leakage occurred. Read that twice. The number on the certificate marks the last step the window survived. It does not mark the step where it broke. The distance between those two steps is real, measurable performance, and it is not in the certificate.
Then three minima are applied before that step-down happens.
Four reductions, then one step down. What you receive is the smallest thing that happened, minus one rung. This is not a criticism of the standards — a class has to be a single reproducible token or it cannot be printed on a label. It is a criticism of reading that token as if it were a measurement.
Here is the most useful line in the whole table. The GB wind-resistance ladder runs from 1 to 9, where level 9 is P3 of 5.0 kPa and above — an open-ended top rung. And the standard adds a note: at level 9, the tested P3 value must be printed alongside the class, in the form level 9 (5.5 kPa). The water ladder does the same thing one rung earlier: at level 5, the tested values for both the fixed part and the opening part must be stated.
That note is an admission. When a rung stops being descriptive, the committee stops trusting it and asks for the number back. The moment the ladder runs out of rungs, the standard asks for exactly the figure the class had been hiding. There is no technical reason to wait until level 9 to request it.
"Performance class" is used to mean at least five unrelated things, and people compare them as if they were the same scale. They are not even the same kind of object.
| What the phrase is being used for | What it actually designates | The ladder | What a "higher" value buys |
|---|---|---|---|
| NAFS Performance Class (R, LC, CW, AW) | Which qualification protocol the product was admitted through | Four letters | A more demanding protocol, not necessarily more pressure |
| NAFS Performance Grade (PG) | The pressure level at which all sub-requirements are met at once | Number, in psf | More design pressure |
| EN 12207 / 12208 / 12210 classes | One class per characteristic, declared on a DoP | Air 1–4; water 1A–9A then Exxx; wind 1–5 then Exxxx, plus deflection A/B/C | Depends entirely on which characteristic you are reading |
| GB/T 7106 grades | Three separate numbers, graded independently | Air 1–8; water 1–6; wind 1–9 on P3 | Three unrelated things; there is no combined grade |
| The casual use: "premium performance class" | Nothing verifiable | No ladder | Nothing |
So "Class 4" is three different quantities depending on which ladder you are standing on. Under EN 12207 it is an air-permeability class. Under EN 12210 it is a wind resistance at a test pressure of 1600 Pa. Under GB/T 7106 it is a wind-resistance grade with P3 between 2.5 and 3.0 kPa. Same numeral, three units of meaning, none of them convertible into the others.
The North American standard had five performance classes in its 2005 edition — R, LC, C, HC and AW. The 2008 and 2011 editions eliminated C and HC and introduced CW, leaving four. More recently, NAFS-26 was published in 2026 while the 2024 editions of the model codes reference an earlier edition. Building codes do not adopt a new edition of a standard the day it appears.
The practical consequence is blunt: a class printed under a superseded edition is a rung on a ladder that may no longer exist. If a quotation shows you a class without an edition year, you have been handed a rung with no ladder under it.
One building-science consultancy put the problem plainly: the performance class does not have a concise description or a precise application, so the only way to understand what it signifies is to read the qualification criteria. Those criteria are published as gateway requirements, and they look like this.
| Class | Minimum PG | Minimum design pressure | Minimum structural test pressure | Minimum water test pressure |
|---|---|---|---|---|
| R | 15 | 720 Pa (15.0 psf) | 1080 Pa (22.5 psf) | 140 Pa (2.90 psf) |
| LC | 25 | 1200 Pa (25.0 psf) | 1800 Pa (37.5 psf) | 180 Pa (3.75 psf) |
| CW | 30 | 1440 Pa (30.0 psf) | 2160 Pa (45.0 psf) | 220 Pa (4.50 psf) |
| AW | 40 | 1920 Pa (40.0 psf) | 2880 Pa (60.0 psf) | 390 Pa (8.00 psf) |
Now read the table the way a buyer usually reads it, as a ranking: R is the weakest, AW is the strongest. Then do one piece of arithmetic. The gateway design pressure for AW is 1920 Pa. A CW product rated at PG80 carries a design pressure of roughly twice that. An AW window sitting at its own gateway is carrying less design pressure than a CW window rated well above its own. Neither product is "better". They were admitted through different doors, and the letter does not tell you which door had the higher pressure behind it.
This is why the trade body that administers the standard draws the distinction explicitly: a high Performance Grade product in a lower class is not the same as a lower-PG product in a higher class. The classes impose different protocols, different deflection limits and different auxiliary test requirements. The letter answers "which protocol". The number answers "how much pressure". You need both, and neither one ranks the other.
The same reasoning that pushes a frequently used double swing door hospital entrance into a heavier duty category also applies here: duty and pressure are separate axes. A window on a ground floor that is opened forty times a day has a different class question from a window on the twentieth floor that is opened four times a year, even if the calculated wind pressure on both is identical.
A class is not a single number that went up. It is a bundle, and the bundle changes shape as the class changes. Several of the items in it are more decisive than the pressure.
| Carried item | What the class does to it | Why it decides your project |
|---|---|---|
| Minimum specimen size | It grows with the class. In one published illustration of the gateway progression for fixed windows, the minimum test specimen runs 1200 x 1200, then 1400 x 1400, then 1500 x 1500, then 1500 x 2500 mm | A class partly tells you how big a window was tested, not how good it is |
| Deflection | In the same illustration, R and LC are "report only"; CW and AW carry a limit of L/175 | At the two lower classes there is no deflection limit at all — the number exists but nothing constrains it |
| Water test pressure | It is derived, not chosen: around 15 percent of design pressure for R, LC and CW, and 20 percent at AW. One North American specifier guide notes a cap for US projects | A water number without the pressure it was derived from is a fraction without a denominator |
| Air leakage limit | Published gateway tables list separate limits for compression seal and sliding seal at the upper classes | The seal principle is written into the table — the same class is a different ask for a casement and a slider |
| Durability and life-cycle testing | Auxiliary durability tests attach at different classes; one specifier checklist associates life-cycle testing with AW | The class determines whether anybody actually cycled the thing before certifying it |
| Operating force and forced entry | Part of the grade set where applicable | A window can hold its pressure and still be unpleasant or insecure to use |
The European wind standard gives frontal deflection three classes: A at 1/150, B at 1/200 and C at 1/300 of the member length, and it combines them with the pressure class into a single designation — C3 means a pressure class of 3 with a deflection of 1/300. That pairing exists because one number was not enough. The GB wind ladder, by contrast, anchors its P1 to a relative deflection of L/300. The North American classes use L/175 once you reach the upper two.
For a 1800 mm mullion, one published industry table spells out what those letters mean in millimetres: roughly 12 mm at A, 9 mm at B and 6 mm at C. The same table notes that national practice differs — one country's technical approvals conventionally require 1/300, a neighbouring country works to 1/200, and another requires 1/300 for residential buildings. The rung you are required to stand on is an administrative decision, not a physical one.
Put the two halves of the process side by side and something uncomfortable appears.
On the requirement side, the design pressure comes from the building — height, exposure category, terrain, position on the facade — calculated under the governing load standard by the project engineer. It is then converted into a performance grade, which is expressed in fixed increments, and the instruction given to specifiers is to round up to the next increment above the calculated value. Nothing may be graded below the number the building asked for.
On the product side, the class is the step below the failure step, and the grade is the lowest of several sub-tests, across three specimens, in the weaker of two directions.
Your real margin is the gap between a number that was rounded up and a number that was rounded down. No drawing, no test report and no label in the submittal package records how wide that gap is.
Three practical consequences follow directly.
Look again at the gateway table. Divide the minimum structural test pressure by the minimum design pressure in each row: 1080 by 720, 1800 by 1200, 2160 by 1440, 2880 by 1920. Every one of them is 1.5. The structural test is already run at one and a half times the design pressure, and the European sequence is explicit about the same structure — a repeated pressure at half the reference pressure, then a safety test at one and a half times it.
This matters because the most common way people misuse a class is to add a class of safety on top of it. Buying a class one rung higher does not add a defined margin; it buys a bigger floor under an unknown ceiling. If you want a known margin, the only thing that gives you one is the tested pressure itself.
A class is not a property of a profile. It is a property of a configuration, and the configuration is assembled from eight things that can each be the weakest.
| Component | What it governs in the test | What it looks like when this is the weak one |
|---|---|---|
| Sash size and aspect ratio | Deflection of the longest member, and the load the hardware sees | The same profile drops a class purely because the window got taller |
| Glass build | Weight, panel stiffness, and the load path into the sash | Maximum sizes are published against an assumed build; change the glass and the size rating is no longer the one you read |
| Mullion and transom depth | Span capacity and stiffness | A shallow depth on a wide span fails on deflection long before it fails on strength |
| Profile wall thickness | Local buckling and fastener holding | A range is printed; the thin member is the one that governs |
| Corner construction | Whether the sash behaves as a frame or as four loose bars | Corner opening under repeated pressure; the air permeability result degrades before anything visibly breaks |
| Locking points and hinges | Whether the compression seal stays compressed around the whole perimeter | Leak at the corners of a tall sash, or a sash that will not stay shut under suction |
| Anchorage and substrate | Nothing, in the laboratory | The specimen is clamped in a steel chamber. Your wall is not a steel chamber, and the class says nothing about it |
| Operating type | Whether the seal can be compressed or can only wipe | A casement class substituted by a slider is not a substitution — it is a different product with a different ceiling |
A casement sash closes by being pulled onto a gasket that runs around the entire perimeter, so the seal is under compression and the locking hardware is what holds it there. A sliding sash has to retain clearance in order to roll, so its seal can only ever wipe against the surface it passes. That is not a quality difference; it is a consequence of how the two things move. It is also why the published gateway tables list separate air-leakage limits for compression seal and sliding seal products at the upper classes — the standard knows the two cannot be asked for the same number.
This is also why the class does not extend by arithmetic. A tested configuration qualifies the sizes at or below it, in the same operating type, with the same glass build. It does not qualify a wider slider, and it does not qualify a 48 folding door simply because the profile is the same family.
We manufacture aluminium system windows and doors. That means we publish a certain set of numbers and not a certain other set, and the honest thing is to show you the boundary rather than blur it.
| What we publish | What you still have to ask for | Why the gap exists |
|---|---|---|
| 6060-T66 alloy across the range | Nothing further | It is the same everywhere, so it is a floor, not a differentiator |
| Profile thickness 1.8–2.0 mm on the C113, C116 and C120 casement systems; 1.8 mm on AW90 | Which member, measured where, coating excluded | It is a range across a section, and the governing member is the thin one |
| PA66GF25 thermal break | The width, in millimetres, on your series | The material name is not a dimension; C116 uses a multi-chamber strip in the 34–44 mm band |
| K of 1.6 on C120, 1.7 on C116, 1.3 on AW90 | A whole-window value for your size and glass | These are system values under stated boundary conditions, and thermal performance is not a performance class anyway |
| 240-hour salt spray and 5,000 N shear, on AW90 | Confirmation that neither is an air, water or wind class | They are material and composite tests on one series, they do not transfer to another, and they measure corrosion and bonding, not weather performance |
| 30,000-cycle hardware figure on the C113 | What the cycle count was measured on | A bench figure for a component is not a service life for a window |
| Frame wall width 90, 113, 116 and 119 mm | Nothing — but do not read it as strength | It is how far the frame goes into the wall, and frame depth is a series property, not a function of how many tracks it has |
| Mullion figures of 75/31, 75/21 and 58/22 | Which of the two numbers controls your span | The first is depth and governs stiffness; the second is the visible face and governs the view |
| Glass sash thickness of 57, 62 and 81 mm | The rebate dimension on the section drawing | Sash thickness is a stack-up, not the glass pocket, and the pocket is what limits your glass build |
| Roller ratings of 200 kg and 400 kg on the sliding systems | The glass build the rating assumes | It is a figure per sash, and the glass dominates it |
| No air, water or wind class on any casement series | A test, or a calculation with its assumptions, against your size, your glass and your required class | A class belongs to a tested configuration. We have not tested yours, so printing one would be inventing it |
That last row is the one that matters most, and it has a hard edge. Deviation is measured against a threshold. If no class is declared on the order, then the word "failed" has no definition, and every leak becomes a negotiation instead of a measurement. Whatever else you do, get a class, an edition and a tested size into the purchase document.
The name on the schedule — a C116 casement, or for that matter a 30 x 80 entry door — is a label for a family of openings. It is not a tested configuration. These are the fourteen lines that turn a label into one.
| Line | Fill in |
|---|---|
| 1. Governing standard, with edition year | ________________ |
| 2. Required design pressure, in Pa | ________________ |
| 3. Who calculated it, from which code and exposure | ________________ |
| 4. Required performance class letter | ________________ |
| 5. Required grade or class number | ________________ |
| 6. Required deflection limit | L / ________ |
| 7. Required water test pressure, and whether derived from design pressure or specified directly | ________________ |
| 8. Required air leakage limit, in the units your code line uses | ________________ |
| 9. Largest unit in the project, width by height, in mm | ____ x ____ |
| 10. Tested specimen size on the offer, in mm | ____ x ____ |
| 11. Glass build the specimen was tested with | ________________ |
| 12. Operating type of the specimen | ________________ |
| 13. Whether durability or life-cycle testing was required, and done | ________________ |
| 14. Pressure actually reached, printed next to the class | ________________ |
If you cannot fill in ten of those fourteen, you do not have a rated window. You have a window with a number printed next to it.
| Ask this | An answer with information in it | An answer that sounds reasonable and carries nothing |
|---|---|---|
| Which class and grade, to which edition? | A full designator with the edition and the operating type | "It meets AW class" |
| What pressure did it actually reach? | A number in Pa, plus the step at which it failed | "It passed comfortably" |
| Which sub-test set the grade? | "Water, at this pressure" | "It is fully compliant" |
| Which direction governed? | "Negative pressure, both were tested" | "It was tested both ways" |
| How many specimens, and was the lowest taken? | "Three, and the lowest governs" | "Tested to the standard" |
| What size was tested, and what is my largest unit? | Both, in millimetres, on the same page | "Standard size" |
| What glass build was the specimen tested with? | The build, and the weight it implies | "Standard double glazing" |
| Is deflection limited, or reported only? | "Limited at this class, reported only at the lower ones" | "It is very stiff" |
| Who signs the field test, and at what pressure? | A named party, and a stated fraction of the laboratory pressure | "The installer will check it" |
| If I change the glass or the size, is it still the same class? | "No. That is a new configuration and needs a new test or a stated calculation" | "The series is tested" |
No. The letter is a protocol and the number is a pressure, and they are different axes. A class above what your building requires buys you a larger mandatory specimen, durability testing and possibly a deflection limit you did not need — while telling you nothing about whether the pressure you actually calculated has been met. Buy the pressure first, then the protocol.
Which ladder? Class 4 under the European air permeability standard is a leakage band. Class 4 under the European wind standard is a test pressure of 1600 Pa. Class 4 under the Chinese wind ladder is a P3 between 2.5 and 3.0 kPa. Three quantities, one numeral, no conversion between them.
Because in one published illustration of the gateway progression, deflection at the two lower classes is report-only, and the water gateway at the lowest class is the smallest of the four. The class people skip when the building is small is exactly the one that was going to constrain how far the mullion bends and how much wind-driven rain the window has to hold back.
Usually a specimen qualifies sizes at or below it, in the same operating type and with the same glass build, but the safe move is to get it in writing with all three conditions attached. And check the direction: if the manufacturer intends to certify larger sizes, the published guidance is that the largest size intended for certification must be the one tested.
No. Certification is issued to an applicant for a tested configuration, and local approval is issued by the authority that has jurisdiction to whoever applies for it. What a factory can supply is the material evidence, the section drawings, the tested specimen and the production consistency behind it. It cannot be the applicant in your country, and a class on our letterhead is not a class your inspector will accept.
It is two different kinds of number. A K value is a thermal property of a system under stated boundary conditions, and we publish it because we have measured it. A performance class is a pressure result on a specimen of a particular size, with a particular glass build, in a particular operating type. We do not print one because we have not tested your configuration, and printing one anyway would be the single most profitable lie available to us.
Send us the fourteen lines from the worksheet, or as many as your engineer has filled in. We will tell you which of them we can answer with a document, which we can answer with a test we would need to run, and which belong to somebody else's signature.
