Private companies to watch
AnthropicIPO coming · Nasdaq: ANTGeneral IntuitionPrivateApptronikPrivateAyar LabsPrivateNuroPrivateStandard BotsPrivatePolymarketPrivate
Live better. Love deeper. Believe bigger.NewsletterSubscribe free
LATEST

SHACL 1.2 Core: What a Passing Data Check Still Leaves to Investigate

W3C published a SHACL 1.2 Core Working Draft on November 16. An original explanation of data-shape checks, useful validation records and questions that require evidence beyond the graph.

A person making written notes beside a desk

By Jonathan Anthony Burkett

Reporting and original analysis | November 17, 2026

A data check can return a reassuring result and still leave an important question unanswered. A record may contain the expected fields while describing an event that never occurred. A date can have the required form while being entered for the wrong transaction. These examples point to a useful distinction in data work: checking whether information fits a stated structure and checking whether it accurately represents the world are related responsibilities, but they are not interchangeable. A new standards draft offers a timely reason to examine that distinction in practical terms.

The World Wide Web Consortium published SHACL 1.2 Core as a Working Draft on November 16, 2026. SHACL, the Shapes Constraint Language, describes constraints on RDF graph structures. The document distinguishes a shapes graph, which expresses the constraints, from a data graph to which they apply. Its status section identifies the text as work in progress and states that Working Draft publication does not imply W3C endorsement. The version date matters: this is a developing specification, not an announcement that every implementation already supports a completed new standard.

Start with a record someone actually needs to use

Consider a hypothetical cultural archive receiving descriptions of objects from several teams. One team supplies a creator, another supplies a place, and a third uses a different label for a date. This is an invented example, not a reported deployment or a test result. Before discussing software, the archive would need to decide what a usable description requires. Which properties matter for its intended task? Which values are optional? Which records are in scope? Those decisions give a validation effort a purpose that a green result alone cannot supply.

A shape can express a requirement that a chosen property have a certain kind of value or occur within a stated count. The draft’s constraint sections describe such mechanisms. But the organization still has to decide which requirement serves its work. A technically precise rule can be poorly chosen for the collection. If an archive treats every object as though it has one known individual creator, the rule may exclude descriptions that honestly preserve uncertainty or collective creation. That would be a problem in the organization’s modeling decision, not proof that the uncertain records have no value.

This suggests a productive first conversation between subject specialists and data specialists. Ask what action a failed check should trigger. Is the record missing information that someone can supply? Does it require an exception? Does the rule misunderstand the material? A useful response depends on knowing which of those situations is present. The conversation should happen before failures accumulate into an anonymous queue. Otherwise, validation can become a process in which people spend time satisfying a form without understanding whether the changes improve the underlying description.

Preserve the question behind the result

A report is more useful when someone can identify the data checked, the rule set used and the version of the process. Those are original workflow recommendations, rather than claims that the specification mandates a particular organizational recordkeeping system. They help a later reader interpret what a result covers. A statement that a collection passed a check is difficult to assess if the collection has changed or the rules are no longer available. Keeping the relevant versions makes the result a traceable event instead of an unqualified badge.

Suppose the hypothetical archive changes its requirement for a date property. A new check may flag records that previously passed. That does not necessarily mean those records became worse overnight. The question may have changed. Staff should be able to distinguish a changed rule from a changed record and a newly discovered error. Without that distinction, a dashboard can create unnecessary confusion. With it, the organization can explain whether it is correcting data, improving its model or testing the effects of a proposed policy before adopting it.

A similar distinction applies to scope. A check may target a defined set of records or relationships. Passing that check says something about the material examined under the chosen constraints. It should not be inflated into a claim about every record in every connected system. In everyday communication, scope can be stated in a sentence beside the result. The reader should be able to tell which collection, period or subset is being discussed without needing to reconstruct the entire project from a technical log.

A valid shape does not supply missing evidence

The draft’s technical sections and examples provide a basis for evaluating the language and its validation mechanisms. They do not remove the need to investigate the source of a claim. In the archive example, a structurally acceptable creator field would still need appropriate evidence for the attribution. A required reference can encourage documentation, but the presence of a reference does not establish that it supports the statement. That last question requires reading and judgment about the source itself.

This is where a data team benefits from allowing different kinds of uncertainty to remain visible. Missing information, disputed attribution and a processing failure are different conditions. An organization can decide how to represent and route them without treating them as interchangeable defects. The model should serve the work people need to do with the collection. A rule that forces an unsupported answer may create a cleaner-looking dataset while making the evidence less honest. That is a reason to review the rule, not a reason to manufacture a value.

For teams following SHACL 1.2 Core, the immediate opportunity is to read the dated draft, examine the relevant constraints and compare them with actual requirements. Any evaluation should identify the software and features tested before making a compatibility claim. This article reports no such test. Its practical argument is about the question surrounding a result: what was checked, under which rules, and what still needs investigation? When those answers remain visible, validation can help people work with evidence more carefully. A passing result then becomes a useful statement with boundaries, rather than a promise no structural check could honestly make.

Image note: Illustrative stock photograph about documenting data rules; not a software interface or evidence of a validation test. Credit: Barbara Olsen / Pexels. Source photograph; Pexels license.

#upstormed #jonathananthonyburkett #jonathanburkett #SHACL #DataQuality

About the author: Jonathan Anthony Burkett is the founder of Upstormed News and Upstormed Inc, an author and songwriter. Meet our team · Our editorial standards · Report a correction

Never miss a story

Get our free Daily Briefing every morning, and follow Upstormed News for breaking news and uplifting stories.

Related stories

Join the conversation

Leave a Reply

Your email address will not be published. Required fields are marked *

Discover more from Upstormed News

Subscribe now to keep reading and get access to the full archive.

Continue reading