Not all dimensions are the same. Vendor name behaves differently from product category, which behaves differently from customer segment, which behaves differently from support ticket status. A practitioner who treats these as a single class will find that the interventions that work for one fail for another. A practitioner who can name the axes along which dimensions differ can choose the right intervention for the right dimension.
This article describes six axes for classifying any dimension. The axes are not the only ones that exist, but they are the six that most reliably predict what a dimension will need from the management discipline. The framework is descriptive: it gives a vocabulary for naming what a specific dimension is, not a procedure for handling all of them.
Axis one: external standard or internal vocabulary.
Some dimensions have governing bodies that maintain a canonical list. ISO country codes. Currency codes. SIC and NAICS industry classifications. UNSPSC product taxonomies. For these, the canonical list exists outside the organisation. The discipline is about adopting and synchronising with the external standard, not creating one.
Most dimensions have no external standard. Vendor names, internal product categories, customer segments, employee roles, deal stages, support ticket types. For these, the canonical list is whatever the organisation chooses to make it. The discipline is about creating and maintaining the list internally.
The implication for management: dimensions with external standards require synchronisation, monitoring for changes to the external standard, and handling of versioning when the standard updates. Dimensions with internal vocabularies require everything that dimensions with external standards require, plus the upstream work of establishing what the canonical values should be in the first place.
Axis two: closed set or open set.
Some dimensions have a finite, knowable set of values. Country (around 200 values). Currency (around 180). US state (50). Months of the year (12). For these, the universe is bounded, and a complete enumeration is possible.
Other dimensions are open-ended. New vendors get onboarded continuously. New job titles get created as the organisation grows. New product categories get added as the catalogue expands. The set grows with the organisation, and enumeration at any point in time is a snapshot of a moving target.
The implication for management: closed-set dimensions can be governed once, with periodic checks for the rare external change. Open-set dimensions require ongoing intake: a workflow for proposing new values, reviewing them, and adding them to the canonical reference. The intake process is where most of the steward's work happens for open-set dimensions.
Axis three: hierarchical or flat.
Some dimensions have natural levels. Geography rolls up from city to state to country to region. Product rolls up from SKU to product line to category to family. Organisational hierarchy rolls up from team to department to function. The dimension is not one column; it is a tree.
Other dimensions are flat. Currency. Status. Severity tier. Deal stage. There is no level structure; each value is a peer of every other value.
The implication for management: hierarchical dimensions require the canonical reference to capture parent-child relationships explicitly, and require validation that prevents impossible combinations (a record cannot have Country: Germany and Region: APAC). Flat dimensions are simpler in structure but require explicit definitions to prevent overlapping or ambiguous categories.
A subtle pattern: dimensions that look flat are often hierarchical in disguise. “Industry” is often presented as a flat list but is actually a two- or three-level tree (Technology > Software > Enterprise Software). When the hierarchy is implicit rather than explicit, the practitioner who needs to aggregate at a higher level discovers that the aggregation is impossible without first reconstructing the tree.
Axis four: stable or evolving.
Some dimensions change rarely. Country codes change every few years when a country renames itself or a new country is recognised. Currency codes change every few years when a currency is introduced or retired. The canonical reference for these dimensions can be updated quarterly or annually with minimal impact.
Other dimensions change continuously. Product taxonomies evolve with every new launch. Customer segments are redefined every time the GTM strategy shifts. Organisational structure restructures with every quarterly reorg. Campaign taxonomies are reinvented every marketing cycle. The canonical reference for these dimensions has to handle change as a normal operating condition, not an exceptional event.
The implication for management: stable dimensions can use a simpler change-management model with longer cycles. Evolving dimensions require the full operating model: stewards available for ongoing decisions, an intake process that handles the change rate, versioning that supports queries across the change boundary, and runtime infrastructure that propagates changes promptly.
Axis five: single-owner or contested.
Some dimensions have a single team that owns them and clearly belongs in their domain. Currency code is treated by Finance. Country code is owned by whoever runs the master location list. Employee role is typically owned by HR. The steward question, “who decides?”, has a clear answer.
Other dimensions are contested. Customer segment is touched by sales (for territory and account assignment), marketing (for campaign targeting), finance (for revenue recognition), product (for prioritisation), and customer success (for tiered service). Each function has legitimate reasons to want segment to behave a specific way. None of them owns it cleanly.
The implication for management: single-owner dimensions can be stewarded by a domain expert with full authority. Contested dimensions require a forum, a clear escalation path, and explicit decisions about which definition wins when they conflict. The forum does not have to be heavy, but it has to exist; otherwise contested dimensions develop multiple parallel definitions that nobody reconciles.
The most consequential decision for a contested dimension is whether to maintain one canonical with subtypes or to acknowledge that the dimension is actually multiple dimensions that happen to share a name. “Mid-Market” as used by sales and “Mid-Market” as used by finance are often different concepts that should be modelled as different fields. The framework helps surface this; the resolution requires organisational judgement.
Axis six: operationally critical or analytically convenient.
Some dimensions are critical for operations. Currency on an invoice determines billing. Country on a customer record determines tax treatment. Status on a support ticket determines routing. Errors in these dimensions have immediate operational consequences and are detected quickly because the operation breaks.
Other dimensions are primarily analytical. Customer segment used for cohort analysis. Industry used for vertical reporting. Deal source used for attribution. Errors in these dimensions accumulate quietly because no operational process breaks; the only consequence is that downstream reports become unreliable, and the unreliability is invisible at the level of any individual record.
The implication for management: operationally-critical dimensions tend to be self-monitoring; the operation surfaces the problem. Analytically-convenient dimensions require explicit monitoring (the drift, resolution coverage, and confidence depth metrics from the target state document) because nothing else will surface the degradation.
A common error: assuming that operationally-critical dimensions are well-managed because they “work.” They work in the sense that the immediate operation succeeds, but the same dimension that successfully routes a single invoice may be entirely useless for analytical questions about invoice patterns. Operational sufficiency is not analytical adequacy.
Reading the framework.
The framework produces a six-tuple for any dimension. Currency, for example: external standard, closed set, flat, stable, single-owner, operationally critical. Vendor name: internal vocabulary, open set, flat, evolving, single-owner (procurement), operationally critical for invoicing and analytically critical for spend analysis. Customer segment: internal vocabulary, open set, sometimes hierarchical, evolving, contested, primarily analytical.
The tuple is not a score; it does not produce a number. It produces a description that determines the management approach. Two dimensions with similar tuples can use similar approaches; two dimensions with different tuples should not be treated the same way.
The most consequential single axis for management approach is axis two (closed vs open set), because it determines whether intake is a primary operational activity. The most consequential single axis for organisational complexity is axis five (single-owner vs contested), because it determines whether stewardship can be straightforward or requires conflict resolution. The most consequential single axis for tooling is axis three (hierarchical vs flat), because hierarchy adds structural requirements that flat dimensions do not have.
A practitioner facing a portfolio of dimensions to manage should classify them all on the framework, group them by the tuples that emerge, and prioritise based on a combination of business criticality (axis six) and management complexity (the interaction of all six). The framework does not produce the priority order; it produces the information from which the order can be reasoned about.
Limits of the framework.
The framework is descriptive, not predictive. It tells a practitioner what a dimension currently is; it does not predict what the dimension will become as the organisation evolves. A closed-set dimension can become open if the organisation expands its definitions. A single-owner dimension can become contested as new functions adopt it. A flat dimension can develop implicit hierarchy as the organisation grows.
The framework is also not the only way to classify dimensions. Other useful axes exist: privacy-sensitive vs not, regulated vs not, customer-facing vs internal, immutable vs editable. The six axes here are the ones that most reliably predict management approach, but a specific organisation may need to extend the framework with axes that matter for its context.
What the framework provides is a starting vocabulary. A practitioner who can say “this dimension is internal-vocabulary, open-set, hierarchical, evolving, contested, and analytically primary” has said more about that dimension than most organisations articulate about any of their dimensions, and has done so in a way that determines what kind of work the management of that dimension will require.