Frequently asked questions

#26: How does the nested format represent entity networks?

Category: Data structure · Last updated: · Permalink

The nested format is a way of including related entities into the description of an entity. For example, a description of a Person entity might include - nested within it's properties - associated Family ties and Sanctions that have been applied to them.

The nested format is returned from the /entities/ API endpoint and is used in the data exports named targets.nested.json and topics.nested.json.

Relationships as entities

Underlying the data model is a simple notion: everything is an entity. To represent relationships in the database, we let entities reference other entities by their ID. For example, a Passport entity has a property called holder that contains the entity ID of the Person to which the document belongs.

In some cases, we want to additionally store data about a relationship. The relationship then is modelled as an entity. For example, an Ownership entity is a link between a Company and Person and can contain information like a startDate, endDate or percentShare. In this case, the Ownership has two properties - asset and owner - that contain the entity IDs of the related entities.

Nesting relationships

The nested entity representation replaces these entity IDs with the full data of the entity they refer to. For example, if a Person has a property addressEntity with an ID value of addr-xxxx, in the nested output this will be a JSON object like this (abridged):

  "id": "individual-1"
  "schema": "Person",
  "properties": {
    "name": "....",
    "addressEntity": [
        "id": "addr-xxxx",
        "schema": "Address",
        "properties": {
           "full": ["New York, USA"]

Inbound relationships

There's an additional twist: entity relationships are interesting when they are outbound (entity A pointing at entity B), but also when they are inbound (entity B being pointed at by entity A).

For example, a Person entity pointing at an Address via its addressEntity property is an outbound reference. A Passport:holder, on the other hand, references the Person it is held by, so the reference is owned by the Passport, not the Person.

In order for that relationship to be included in a nested Person entity, we use a stub property: in this case, Person:identification will contain an array of Passport entities, all of which have the Person's ID as their holder property. Stub properties are used to assign a property name to the opposite direction viewpoint of an entity reference.

Another example: the Ownership entity references an owner, for example a Person or Company. In a nested data extract, that Person or Company would include a list of Ownerships in its ownershipOwner stub property, which is the inverse of the Ownership:owner property.

Nesting depth

In theory, nested entities could contain an infinite depth of nested data. However, both the API endpoint and the data exports limit the depth to one step by default: a Person will contain a list of Address in addressEntity, but those Address entities will not, in turn, nest all the things located there ( things is the stub property inverse of addressEntity).

There's an exception to this, for entity types like Ownership. They exist to connect two other entities (in the schema metadata, they are marked schema.edge). Such edge entities will contain a further level of nesting. This way, the asset owned by a Person via an Ownership is included in the nested representation of the Person as well.

Once the nesting depth is exhausted, the data will revert to including the IDs of referenced entities instead of expanding them into entities. This allows you to manually recurse into more depth by looking up the relevant entity.

Related questions

« Back to full FAQ index