The discipline, written down.
Five foundational documents that argue and specify the practice. A catalogue that names every failure mode by layer. A growing library of practitioner artifacts. An interactive diagnostic that walks from symptom to root cause. Everything here is open, citable, and authored to be read carefully.
Four essays argue the case. One catalogue indexes the failure modes.
The five documents are functional peers: foundational to the discipline. They are not stylistic peers. The essays read front-to-back. The catalogue is consulted by section.
- 01ESSAY · FOUNDATIONAL
Manifesto
The categorical surface, governed in production.
Argues why dimensional inconsistency, tolerable for thirty years, becomes intolerable in the agent era. The motivating document for the whole pillar.
~ 12 minRead - 02ESSAY · FOUNDATIONAL
Primer
Dimensional data, established from first principles.
Establishes the category for a reader new to dimensional data. Observational rather than argumentative; pairs with the narrative origin-story article in the library.
~ 18 minRead - 03ESSAY · FOUNDATIONAL
Target state
What good looks like, specified.
Specifies the discipline at the level of the canonical reference, the runtime, the workflow, and the metrics that measure whether the system is working.
~ 22 minRead - 04ESSAY · FOUNDATIONAL
Origins
Why dimension management emerged as a discipline now.
Traces the conversational-analytics shift that removed the human reviewer from the last mile, the upstream redistribution of quality control, and the structural gap that dimension management exists to fill.
~ 18 minRead
The dimensional data problem
Four layers. Thirty-four named failure modes. Citable, deep-linked.
A structured reference catalogue, not an essay. Layer 1 catalogues eight categories of data-layer variance, Layer 2 nine categories of process-layer breakdown, Layer 3 eight organisational root causes, Layer 4 nine categories of downstream business harm. Practitioners cite individual entries by anchor.
Artifacts. Specific topics, expanded.
The library expands on topics the foundational documents introduce. Two articles published before the v2 framing exist and are being edited for republication; four new artifacts are drafting. The library grows when the substance is ready, not on a content-marketing cadence.
- 01ARTICLE · ORIGIN STORY
Why Dimensional Data Gets Messy
How inconsistent categorical data forms, why it persists, why it is not anyone's fault. Narrative companion to the primer.
~ 9 min - 02ARTICLE · POSITIONING
Why Dimensional Data Outlives Every Tool
The DQ-vs-MDM positioning. What canonical means, where each adjacent category lands, and the three structural reasons the problem persists.
~ 11 min - 03ARTIFACT · MODEL
Variance evolution: the four phases
How dimensional variance accumulates over time and what to look for at each phase. Pairs with a visualisation.
~ 7 min - 04ARTIFACT · FRAMEWORK
Six axes for classifying a dimension
A practitioner framework for deciding which dimensions warrant active management and which can be left to the system.
~ 8 min - 05ARTIFACT · REFERENCE
What counts as a dimension
A reframing of the data-types universe around the question of which kinds of data are dimensional. Companion to the primer.
~ 6 min - 06ARTIFACT · BUYER-FACING
The cost of leaving dimensions unmanaged
A buyer-facing condensation of Layer 4 of the problem taxonomy: the downstream consequences in business terms.
~ 5 min
Tools that test the framing against your own experience.
Trace the Chain
The four-layer taxonomy, made interactive.
Walk from a recognisable symptom, a report that doesn't add up, analysts spending days on cleanup, dashboards being abandoned, down through the data layer, the process layer, and the organisational layer to a root cause. Runs in reverse too: from a suspected root cause to the impacts it produces.
Open the diagnosticDimension management, in brief.
Common questions, condensed to a paragraph each. For depth, the foundational documents above and the glossary in Learn are the authoritative sources.
What is dimension management?
Dimension management is the discipline of governing the categorical surface of business systems: the labels, classifications, segments, statuses, categories, and types that turn raw transactions into analysable populations. It is distinct from master data management, which resolves entities; dimension management governs the values those entities are described by.
What is a canonical reference?
A canonical reference is the authoritative source for a dimension. It holds the accepted values, their definitions, ownership, version history, and aliases. The reference is system-agnostic: it does not live inside a warehouse or catalogue or master-data system; it is its own surface, exposed via API to every system that produces or consumes the dimension.
How is dimension management different from master data management (MDM)?
MDM answers the question "are these the same entity"; dimension management answers "what segment is this entity in, and what does that segment mean". MDM operates at the entity layer; dimension management operates at the value layer that the entity attributes are drawn from. The categorical attributes on master records get populated by survivorship rules, which have no view on whether the surviving value is canonical.
How is dimension management different from data observability?
Observability detects volume, freshness, and distribution drift. It covers structural and statistical quality control well. Dimension management governs categorical semantics: variant resolution, canonical definitions, naming, and hierarchy decisions on categorical values. A monitor can tell you that a new value appeared in the vendor field; it cannot tell you what that value should resolve to.
Why is dimension management emerging as a discipline now?
The shift from dashboards to conversational analytics has removed the human reviewer who was, for thirty years, the unwritten quality gate at the last mile of analytics. Without that reviewer, the reactive cleanup model breaks. Agents produce categorical values at machine speed; configuration variance compounds at the same speed. Reactive cleanup, which worked when humans were the dominant producers, is about to become structurally unsustainable.
What are the five forms of dimensional variance?
Surface variance (same value, different formatting: AWS, aws, A.W.S.). Semantic variance (different labels for the same concept: vendor, supplier, partner). Definitional variance (same label, different populations across teams: Mid-Market means different things in finance and sales). Granularity variance (same concept at different hierarchy levels: Financial Services, Retail Banking, BFSI). Temporal variance (canonical form changes over time, historical records keep the old form).
What is CleanDims?
CleanDims is dimension management infrastructure: a canonical reference, a runtime, and a workflow. Three surfaces, federated by domain owner. The reference holds the accepted values, definitions, aliases, and version history per dimension. The runtime is the propagation layer that agents and pipelines read from. The workflow is the request queue stewards work through. Everything CleanDims publishes about the discipline is open and citable.
The practice belongs to the category, not to any one product.
Everything in this pillar is published under CC BY 4.0. Cite it, quote it, build on it. The expectation is attribution, not permission. Each foundational document and library entry carries a citation block with a suggested citation format.
The library is small on purpose. New artifacts ship when the substance is ready and not before. If you are working on something that belongs in the library, have noticed a failure mode that the four-layer taxonomy does not cover, or have a correction to anything published, send it through the contact form. One channel, one form, for every kind of inquiry.
NOT IN V2A public contribution platform, a comments system, or a discussion forum. These can come later if there is demand. v2 contribution is a direct line to the people writing the documents.