Origins.
Eight pages, read in sequence. About thirty-five minutes. Why dimension management emerged as a discipline now.
The Foundations module establishes what dimensional data is and what well-managed dimensional data looks like. The Origins module asks a different question. It traces the shift in analytical architecture, dashboards being replaced by conversational interfaces, that turns dimension management from a quietly tolerated tax into an explicit discipline.
The argument runs in eight pages. The first three describe the shift itself: the control that lived between the dashboard and the executive, the move to conversational analytics, and what that move removes from the pipeline. Pages four and five trace the consequence: without the last-mile reviewer, quality control has to move upstream and distribute across four layers. Page six explains why one of those four layers, dimension management, requires humans in the loop where the others can run on automation alone. Pages seven and eight close on the structural argument: why no existing tool category absorbs the work, and why the discipline has to sit above the stack rather than inside any one tool.
The module is written for practitioners and leaders who want to understand why this discipline is appearing now. Foundations is recommended but not required; pages cross-reference the glossary throughout for terms introduced there.
The eight pages.
- 1
The control that was never written down
How the dashboard model relied on the analyst as an unspecified quality gate, and why the gate worked at all.
~ 5 min - 2
The shift to conversational analytics
The artifact moves from the dashboard to the agent answer. What that means for the pipeline upstream and the reader downstream.
~ 4 min - 3
What disappears with the dashboard
Two pieces of the old model do not survive the shift: the dashboard as artifact, and the human review at the last mile.
~ 4 min - 4
The consequence: no safety net
Agent answers reach the executive without a buffer. The three failure profiles that produce wrong charts without raising errors.
~ 5 min - 5
Quality control moves upstream
The check that used to happen at the visualisation layer distributes across four upstream layers: contracts, observability, dimensions, metrics.
~ 5 min - 6
Why dimensions specifically need humans
Of the five forms variance takes, only one is safely automatable. The other four require an organisational decision, not an algorithmic one.
~ 5 min - 7
Why existing tools fall short
Observability, MDM, catalogues, and semantic layers each touch the categorical surface. None treats it as the primary object.
~ 5 min - 8
Why horizontal, not in-house
A reference embedded in any one tool is unreachable to the others. The discipline has to sit above the stack, not inside any of its layers.
~ 5 min