Gap Catalogue: Connections That Should Exist¶

Type: Non-normative

The metamodel shows the connections between data contracts, SMF documents, implementations and decisions. A gap is a connection that should exist and does not: a contract with no owner, a build with no contract behind it, a claim resting on a build nobody has tested.

Gaps are the work list. They are also the honest measure of how far Semantic Management has been adopted, because each one belongs to a practice, and a practice is in place when its gaps are closed for the terms that matter.

This catalogue names the gaps so tools can find them and people can talk about them. It does not say every gap must be closed. Govern in proportion: close the gaps on the terms where confusion, reuse, AI use, regulatory exposure or executive visibility make it matter, and leave the rest open on purpose.

How to read a gap¶


Name	What is missing, in plain words
Looks like	How it shows up in the organization
Practice	Which SMF practice owns closing it
Weight	How much it matters when the term matters: critical (a wrong number or a guess is likely), high (a dispute or a silent change is likely), medium (inconvenient, not dangerous)

Meaning¶

Name	Looks like	Practice	Weight
Term with no concept	A word people use in reports and questions that nobody has written down. Everything built on it is a guess	Discovery, Modeling	high
Concept with no owner	A definition exists but nobody can approve a change to it	Authority	critical
Perspective with no owner	"Service gross margin" exists as a label but nobody is accountable for what it means	Authority	critical
Two meanings, one label, no decision	The same term is used two ways and nobody has said whether they are the same, both legitimate, one inside the other, or an open conflict	Authority	critical
Conflict past its date	A disagreement was declared with an owner and a target date, and the date has passed	Authority, Lifecycle	high
Retired term with no replacement	Something was deprecated but the thing to use instead is not named	Lifecycle	high
Concept with no classification rule	A term like "active customer" where what counts has never been written down, so every build decides for itself	Modeling, Authority	high

Resolution¶

Name	Looks like	Practice	Weight
Several perspectives, no answer rule	A concept has two or more legitimate meanings and nothing says which applies where. Every consumer picks one silently	Resolution	critical
Answer rule with no "ask"	The rule names a perspective for some contexts and falls back to a default for everything else, so an unlisted situation gets a guess instead of a question	Resolution	critical
Context nobody defined	Answer rules refer to a situation ("board", "leadership meeting") that is not written down anywhere, so tools cannot recognize it	Resolution	medium
Enterprise default used as a fallback	The default meant for board and external reporting is being served to every unmatched question	Resolution, Authority	high

Measurement¶

Name	Looks like	Practice	Weight
Perspective with no Metric Contract	A legitimate meaning of a number, with an owner, but no agreed definition of how it is measured	Measurement	high
Metric Contract with no owner or sign-off	A definition exists but nobody has approved it	Measurement, Authority	critical
Metric Contract with no approved uses	The number is defined but it is not written down where it may and may not be used, so it ends up in the board deck	Measurement	high
Contract pair with no comparability statement	Two contracts measure the same concept and neither says whether they may be compared or combined. Someone will average them	Measurement	critical
Metric Contract with no data contracts beneath it	The number is defined but it is not recorded what datasets it stands on, so a change upstream is a surprise	Measurement	medium
Metric Contract with no version or effective date	The definition has changed and last year's numbers can no longer be read under the definition that produced them	Lifecycle	high

Implementation¶

Name	Looks like	Practice	Weight
Metric Contract with no build record	The agreement exists; nobody has recorded where it is actually calculated	Implementation	high
Build with no contract (shadow metric)	A measure in a BI model, a semantic layer or an interchange model that no contract governs	Discovery, Implementation	high
Build never tested	A build record exists but its conformance has never been assessed	Implementation, Assurance	high
Build with no snapshot	The build was assessed, but the calculation as it was at that moment was not kept, so drift can only be noticed by chance	Implementation	medium
Build has drifted	The calculation in the tool no longer matches the snapshot on its build record	Implementation, Assurance	critical
Generated build with no source	A semantic-layer or BI measure was produced from an interchange model, but the build record does not say so, so it gets tested as if it were hand-built	Implementation	medium
Interchange model with no pointer back	An Ossie metric, a semantic layer metric, a data product or a data contract field is governed by a contract but does not say so in its own file, so a tool reading only that file cannot find the agreement	Interchange	medium

Consumption and assurance¶

Name	Looks like	Practice	Weight
Consumer with no checks	A report suite or AI assistant uses governed terms and nobody has written a single question with an expected answer	Assurance	critical
"Ask" case never tested	The answer rule says a situation should trigger a question, and no check confirms the consumer actually asks	Assurance	critical
Check never run	Checks exist; there are no results, or the results predate the last definition change	Assurance	high
Consumer that guessed	A check expected "ask" and the consumer answered with a number	Assurance	critical
Claim with no contract version	A number in a board deck or a filing that cannot be traced to the agreement that defined it	Assurance	critical
Claim on an untested or drifted build	The number is traced to a build whose conformance is untested, non-conformant or stale	Assurance	critical
Claim with no verification	The trace exists but nobody has confirmed the number by re-execution or reconciliation	Assurance	high

Lifecycle¶

Name	Looks like	Practice	Weight
Change with no note	A definition's version went up and nothing says what changed, why, or from when	Lifecycle	high
Superseded definition still in use	A build or a claim still points at a retired version with no replacement recorded	Lifecycle, Implementation	high

Using the catalogue¶

As a work list. Run it over the terms that matter and sort by weight. The critical gaps on executive metrics are the first quarter's work.
As a maturity reading. A practice is in place for a term when its gaps are closed for that term. Counting closed gaps per practice, over the governed terms, gives a maturity picture that is observed rather than self-assessed. This is the basis for the maturity model on the roadmap.
As tool behavior. A tool that renders the metamodel can show each gap as a missing connection on the card it belongs to. Most gaps are mechanical to find from the documents alone; a few (drift, checks never run) need the tool to read the implementation or the results.
As AI governance evidence. The assurance gaps map directly onto the Measure and Manage functions in ai-governance-evidence.md.

Names and weights here are a first cut from practice. Challenge them, add missing gaps, and bring cases where a gap was harmless or a closed gap still produced a wrong number. See CONTRIBUTING.md.