Why dimensions stay unmanaged.

The first condition is that dimensions are owned by no one. The vendor name field is touched by procurement, finance, accounts payable, the data team, and at least one external system integrator. Each function uses the field. Each function modifies the field. No function is accountable for the field as a whole.

When a vendor name is wrong, the question of who should fix it is itself contested. Procurement says the data team should clean it in the warehouse. The data team says procurement should enforce it in the source system. The source system administrator says it is a process problem for procurement to solve. The conversation never resolves because no one has the authority to decide who decides.

This is not a procurement-specific story. The same pattern applies to product categories, customer segments, support ticket types, deal stages, and every other dimension where multiple teams produce and consume the values. Distributed ownership is the default; centralised ownership is rare; explicit named ownership of the kind the target state describes is rarer still.

The second condition is that the existing governance disciplines do not cover the categorical surface. Master Data Management was built to govern entities (the customer, the product, the supplier as objects) and resolves which record is the authoritative golden record for that entity. It does not resolve the categorical attributes on the entity record: which industry label the customer carries, which category the product belongs to, which classification applies to the supplier.

Data Quality tooling was built to validate that data conforms to rules. It catches structural problems: nulls, type mismatches, range violations, referential integrity failures. It does not catch semantic equivalence problems where two valid strings refer to the same concept in the world.

Data Catalog organises metadata about data assets and helps discoverability. It does not govern the vocabularies that classify and describe those assets.

The result is that adjacent disciplines each address some part of the data management problem and none addresses the categorical surface specifically. The longer argument for this is in Why Dimensional Data Outlives Every Tool.

The third condition is that dimensional inconsistency does not surface as an error. A numeric field with a wrong value triggers an investigation because the wrong total is noticed. A categorical field with eight variants of the same value produces a result that looks plausible, because the variants individually all appear valid. The error surfaces weeks or months later, in a meeting between two leaders looking at two dashboards built on the same warehouse, and the discrepancy gets attributed to “we need to reconcile our reporting” rather than to the dimensional layer.

This invisibility is structural. The variance is in the data; the consequences are in the analyses; the connection between them is rarely traced. By the time the connection is made, the variance has been accumulating for months or years.

The fourth condition is that dimension management is unglamorous work. Standardising vendor names builds nobody's career. It produces no new capability, adds no feature, and generates no metric anyone celebrates. The consequence of doing it badly is borne by other people on a different time horizon. The consequence of doing it well is invisible, because well-managed dimensions produce reports that simply work.

Organisations systematically underinvest in work with this profile. The work that gets funded is the work that has visible outcomes; dimension management has invisible outcomes by definition, because its job is to make a category of failure stop happening rather than to produce a new capability.

The four conditions reinforce each other. The ownership vacuum prevents anyone from championing the work. The lack of governance fit prevents the existing tools from solving the problem. The invisibility prevents the cost from being measured. The unglamorous nature prevents the work from being funded. The same organisation that has invested heavily in data infrastructure has not addressed the categorical surface, because the conditions that produced the gap continue to produce the gap.

What changes this is not better individual decisions. It is the recognition that dimension management is a discipline in its own right, requiring its own ownership, its own governance, its own metrics, and its own funding. That recognition is the argument the manifesto makes, the discipline the target state describes, and the work the rest of CleanDims exists to support.

The detailed taxonomy of organisational conditions that allow dimensional problems to persist is in Layer 3 of the problem catalogue, which enumerates eight specific conditions.