# I Scored 532 Data Model Standards on Whether They Modernized

**Published:** 2026-08-20  
**Author:** Kin Lane  
**Canonical:** https://apievangelist.com/2026/08/20/i-scored-532-data-model-standards-on-whether-they-modernized/

There is a layer underneath every API I profile that I have never been able to point at as a whole. HTTP and OpenAPI describe how you move data. FHIR, ISO 20022, IEC CIM, NIEM, and a few hundred others describe what the data actually *is* — the entities, the attributes, the relationships, and the controlled values that fill them. That layer is older than the API space, it is stewarded by different bodies, and until recently there was no way to ask a catalog to show it to you. So I built one. There are now [532 data model standards](https://standards.apievangelist.com/models/) published at standards.apievangelist.com, across 139 domains, each one fetched and confirmed against its steward's own site before it was allowed onto the page. Then I scored 530 of them on a single question: **is this standard of the current era, or is it stuck in the one it was born in?**

I had to define what "modernizing" means before I could measure it, because the word gets used to mean everything and therefore nothing. I settled on eight things I could actually observe from the outside: whether the serialization is current, whether the artifact is retrievable, what the access economics are, whether there is an API surface, whether development happens in the open, whether change is published, whether conformance tooling exists, and whether any of it is legible to an agent. Those eight roll up into a modernity band. Crucially, I kept access as a *separate* axis rather than folding it into the band, because enclosure is not a point on the modernity scale — SNOMED CT is technically current and licensed content at the same time, and collapsing those two facts hides the one you needed. Every standard reports as a pair, like "Modernizing / Enclosed."

The serialization numbers are the most useful thing I found, because they show what modernization looks like in practice rather than in a strategy deck. Of the 530 I scored, 128 are modern-native — JSON Schema, OpenAPI, Protobuf, or plain JSON and YAML named as a form of the model. 69 are semantic-native, published as RDF, OWL, SKOS, or queryable over SPARQL. 64 are XML and XSD only. Two are still EDI or fixed-format only. And 81 sit in the middle: an XML-era core with a JSON or RDF binding named alongside it. **That group of 81 is the answer.** They did not throw away decades of consensus and edge cases to chase a format. They kept the normative model exactly where it was and added a binding that today's tooling can pick up. I wrote about this instinct qualitatively back in July — [the work is re-expression, not reinvention](https://apievangelist.com/2026/07/04/bringing-legacy-standards-into-the-modern-api-age/) — and it is satisfying to now be able to say that 81 standards have actually done it, and to name them.

The sharpest line in the data is not about format at all, it is about access. The median score for standards you can freely get is 66.2. For enclosed standards — the ones where the specification is sold or the content is licensed — the median is 38.9. Standards behind a partial gate land at 45.9. And here is the number I keep coming back to: of the 62 standards that earned the Current band, **not one of them is enclosed**. Fifty-eight are open and the other four have an access posture I have not established yet. Zero out of 62. Meanwhile 32 of the 61 Legacy standards are enclosed, and only five are open. I did not design the rubric to produce that result and I would not have predicted it was quite so clean. Modernization turns out to be a distributed act — it happens when implementers, tool authors, and CI pipelines can all reach the artifact and build against it. Put the model behind a purchase order and there is nobody on the other side to do the modernizing.

Then there is the finding that should worry anyone shipping an agent strategy. Agent legibility is worth five points out of a hundred and asks something modest: is there an `llms.txt`, is there a persistent identifier scheme, is there anything here an automated client can orient on? **419 of 530 data model standards score zero on it.** Not a low score — a zero. The layer that defines what a payment is, what a patient encounter is, what a grid asset is, what a shipment is, is very nearly invisible to the agents that everyone is now pointing at their enterprise data. We spent the last two years making the API surface agent-ready, and this is what sits underneath it. That is a five-point line item on a rubric and probably the biggest single gap in the whole catalog.

I want to be careful about what a low score means here, because it would be easy to read this as a ranking of worth and it is not. EDIFACT, ASC X12, and SWIFT MT move an enormous share of global commerce every day and they sit low on this rubric by design, because the rubric measures era and openness, not value or reliability. A closed-era standard that clears trillions is not a failure, it is a fact about where the modernizing work has not happened yet — and often for very good reasons involving the cost of moving an installed base. I also want to be honest about coverage: 226 of the 530 come back unassessed rather than banded, because I hold anything under 60% evidence coverage out of the bands entirely. Letting thin evidence fall through to the bottom band would turn my own blindness into an accusation against a steward, which is a defect I have chased out of my scoring work before and do not intend to reintroduce here.

If you steward one of these, the path off the bottom is shorter than it looks and it is not a rewrite. Put the schema artifact at a stable, fetchable URL that does not require a login, so a pipeline can pull it. Add a modern binding beside the normative source rather than forking it, the way those 81 dual-form standards did. Publish a changelog so downstream implementers can see what moved between versions. And spend an afternoon on agent legibility, because at the moment that is a nearly empty column and being one of the few standards in your domain that an agent can actually orient on is worth more right now than it will ever be again. The catalog is [open at standards.apievangelist.com/models/](https://standards.apievangelist.com/models/), it filters by layer, domain, access, and modernity band, and if your standard is in there and I have it wrong, tell me and I will fix it.