Published 20 August 2026. Every count on this page is produced by www/content/tools/audit.py
from the data files, not typed. If a figure here and that script’s output ever disagree, this
page is the one that is wrong.
What is actually there
| Grades | 27 |
| Sourced values | 191 |
| Named gaps | 82 |
| Values above E3 | 0 |
| Documents supplying all of it | 8 |
Two values are claimed at E5 in the data files — the ASTM proof strengths for 316L and 2205 — and both are demoted to E3 automatically, because the standard behind them is paywalled and nobody here has opened it. That demotion is applied by the generator, not by an editor remembering to be modest. How the ladder works.
So the honest summary of this site’s data layer is: everything on it is a real value from a resolvable supplier document, and nothing on it is verified against the standard it cites. That is E3, and E3 is a screening level. None of it is a design allowable.
The dependency we would rather not have
Of 191 values, 188 come from Outokumpu range datasheets — Core, Supra, Forta, Ultra, Therma and Dura. All six carry the same edition date: May 2015. They were eleven years old when we retrieved them.
The remaining three are one distributor datasheet and one secondary comparison page.
This is a concentration risk and it is worth stating in a whole paragraph rather than a footnote. If one of those six documents contains a transcription error, a superseded figure or a quietly revised minimum, it does not affect one page here — it affects a family. The PREN column is checked arithmetically against each datasheet’s own stated composition and formula, which catches mis-transcription and caught nothing, but that check cannot tell us whether the document itself is current. A 2026 revision of any of these datasheets could move numbers we publish, and we would not know until someone looked.
The fix is not clever. It is a second independent source per grade, and we have it on exactly one: 441, where a distributor datasheet turned out to disagree with nothing and explain something instead.
Where the 82 gaps come from
| Reason | Gaps |
|---|---|
| The source is silent — no such column, no such row, an empty cell | 39 |
| The standard is paywalled and was not opened | 31 |
| Other, including product forms the datasheet does not list | 12 |
The single most common gap is the same one, 23 times: composition limits.
That is worth sitting with. On 23 of 27 grades we can tell you the minimum proof strength but not what the alloy is permitted to contain. The datasheets print typical compositions — useful enough that we recompute every PREN from them and they reproduce — but a typical composition is not what a purchase order specifies, and treating one as the other is the same error this site spends most of its time arguing against. Specified against typical.
Composition limits live in ASTM A240 and EN 10088-2. Both are paywalled. That is the whole reason for the largest hole in our data, and it is not a hole we can dig out of by finding a better free source, because a free copy of a paid standard is not provenance either.
What would move these numbers
Two documents, opened once, would change most of the table above.
ASTM A240/A240M. Promotes the two E5-claimed values from E3 to E5 with an edition and a table, and supplies ASTM composition limits and plate minima across the austenitic and duplex grades. It is the highest-value single purchase available to this project.
EN 10088-2. 63 of our values are attributed to it through a datasheet. None of them can rise above E3 while the standard itself is unread, no matter how good the datasheet is — that is what the demotion rule is for, and applying it to ourselves is the only reason anyone should believe it when we apply it elsewhere.
Six further gaps rest on EN 10095, behind the two heat-resistant grades whose elevated temperature rows are the only numbers those pages carry at all.
Why publish the audit
Because the alternative is to publish 191 numbers with a confident tone and let the reader discover the shape of the uncertainty by accident, which is what the rest of the field does and is the reason this site exists.
There is also a narrower argument. A gap with a stated reason is a usable object: it tells a
reader whether the missing value is missing because nobody measured it, because the document was
silent, or because the document costs money. Those three lead to different next steps. An empty
cell leads to none. That is the rule the generator enforces — a property marked
not established without a reason will not build.
We publish the count because the count is the strongest claim we can make. Anyone can say their data is traceable. We can say how far, and where it stops.
It is also easier to publish an unflattering audit when there is one name on the site rather than a committee. Nobody had to be persuaded.
What this note does not cover
- Whether the 191 values are correct. This is an audit of provenance, not of accuracy. The arithmetic check on PREN passes on all 21 grades that have one; nothing else here is independently verified.
- Whether the May 2015 datasheets have been superseded. We know they are old. We have not established what a current revision says.
- The editorial pages. The counts cover grade and comparison pages generated from data files. The written pages carry their own figures with their own sources.
- Our own conformance to the framework we publish. That has a separate page, and it does not currently pass at Level 1.
- What a good number would look like. A mill certificate for your actual heat reaches E5 and supersedes everything here. We do not have your heat.