A property is a feature, attribute or characteristic assigned to an object, event, system, person, place or other entity.
Colour, mass, ownership, location, temperature, age, confidence, capacity and risk level can all be properties. But they are not the same kind of property: some are intrinsic, some relational, some observed directly, some derived, and some change with time or context.
Quick answer: how should properties be categorised?
- Source: observed, measured, asserted, inferred or computed?
- Form: intrinsic or relational?
- Stability: fixed, slowly changing, dynamic or event-driven?
- Context: universal or dependent on place, role, time or viewpoint?
- Value type: categorical, numeric, ordinal, Boolean, text or structured?
- Derivation: raw or calculated from other properties?
- Confidence: certain, estimated, disputed or missing?
- Unit: what measurement system applies?
- Scope: does the property describe the whole entity or one component?
- Version: when was the value valid?
This page complements How to Categorise Objects and How Ontologies Work by treating properties themselves as first-class objects.
Property is not identity
An entity can change many properties while remaining the same entity. A car can change colour, owner and location without becoming a different car.
Intrinsic properties belong to the entity itself
Mass, material or serial number can be treated as intrinsic within a given model, though even these may change over time.
Relational properties depend on another entity
Owned-by, located-near and parent-of cannot be understood without the relation target. They are often better represented as relationships than simple scalar fields.
Observed properties are recorded directly
A visible colour or stated status may be captured from direct observation, though observer and context still matter.
Measured properties use a method and unit
Temperature, length and latency require measurement provenance, units and uncertainty.
Derived properties are calculated
Speed from distance and time, age from birth date, and utilisation from capacity and load are examples. Preserve the formula and source values.
Inferred properties are model-dependent
Risk level, intent and likely state may be inferred from evidence rather than observed directly. Confidence and method should travel with the value.
Stable properties change slowly
Material composition or birth date may remain fixed for practical purposes, while address, ownership and configuration change more often.
Dynamic properties require time
Temperature, stock level, location and status are incomplete without a timestamp or validity period.
Event-driven properties change at discrete transitions
Approval state, membership and ownership often change when a defined event occurs rather than continuously.
Contextual properties depend on viewpoint
Priority, relevance, accessibility and risk can differ by user, task, jurisdiction or scenario.
Categorical properties use controlled labels
Material type, status and region can use enumerated categories. Controlled vocabularies reduce spelling and synonym drift.
Numeric properties need scale and unit
A bare number is ambiguous. Quantity, unit, precision and uncertainty determine meaning.
Ordinal properties preserve order without equal distance
Low, medium and high convey ranking but not necessarily equal intervals.
Boolean properties compress complexity
True/false flags are useful operationally, but they can hide unknown, not-applicable or contested states if those are not modelled separately.
Missing is not false
Unknown, not observed, not applicable, withheld and truly absent should remain distinct where the difference affects reasoning.
Property provenance matters
Record who or what supplied the value, when, by which method and under which version or authority.
Conflicting property values need local uncertainty
Two sources can disagree about ownership, age or capacity. Preserve competing values and evidence rather than silently choosing one without justification.
Property drift can reveal system change
A shift in average latency, risk or demand may be real, or may come from changed measurement definitions. Version both the property value and the method where needed.
A practical property record
- property ID and name;
- entity described;
- property type;
- value and value type;
- unit or vocabulary;
- intrinsic or relational status;
- observed, measured, inferred or derived source;
- formula if derived;
- context;
- valid time;
- uncertainty or confidence;
- provenance;
- version.
The deeper idea
Properties are how knowledge systems turn entities into describable structures. Their usefulness depends on keeping value, context, time and provenance attached.
To categorise a property well is to know what kind of characteristic it is, how the value was obtained, whether it depends on context or another entity, and when that value was valid.
Final answer
Categorise properties by source, intrinsic or relational form, stability, context, value type, derivation, confidence, unit, scope and version. Keep properties separate from identity and relationships where appropriate, and preserve time and provenance for values that can change or be disputed.