Skip to main content

Glossary

(PE) indicates a term associated with Portrait Engine.

Account. An account is a named LiveRamp relationship that can be used to log into an application or service.

Attributes. Attributes the fields that are output in the first party graph. These can be passed through from the source files (Passthrough Attribute) or calculated across source files (Calculated Attribute),

Batch data. Collections of data that are processed as a single operation.

Blocker. A blocker is a candidate identifier for matching. Records are grouped into pairs as candidates for matching based on the presence of a shared blocking key. (PE)

Consolidations. A consolidation merges two or more identities into one.

Data source. A data source is a file in a specific format located in a specific location.

Dataset. A dataset is a named collection of data.

Deployment. A collection of services that run in the cloud to support certain functionality.

Dimension. A dimension is a measurement of data and relationships that separates one identity from another.

Domain. A domain is the customer-specific encoding applied to Portrait IDs, which makes IDs that represent the same entity distinct for each client. A client domain can be identified in the first 4 digits of the link format.

Entity Resolution. Entity resolution is a technique to identify data records in a single data source or across multiple data sources that refer to the same real-world entity and to link the records together. The more specific version of entity resolution that deals with humans is called identity resolution. The goal of entity resolution is to create a common identifier across datasets where there is none.

Filter. A rule(s) for selecting a subset of data. In Portrait Engine, a filter is a function that removes rows from the output based on criteria that a client creates in the filter extension using Boolean logic. For example, a client could configure a filter that would prevent any Portrait ID with a value of “TRUE” in the “OPT OUT” attribute of the portrait from being present in the output file.

First-Party Data. First-party data are data that is owned or gathered by the client with full rights to use for marketing, analytics, or other use cases. Since the party in question gathered the data, the privacy policy for the company would apply to that data, and notice and consent will have been provided as required. (PE)

First Party Graph (PE Output): A first-party graph is an identity graph that resolves entities to person, household, or other resolution levels using first-party data only. In Portrait Engine, clients are also able to specify and change the resolution methodology and rules used in building the graph. (PE)

Graph generation: A Graph Generation is a version of a graph that is generated by a Portrait Engine application running. Each application will have several graph generations, correlating to each time that an application runs.

Graph Generation. A Graph generation is a version of a graph that is generated by a Portrait Engine application running. Each application will have a number of graph generations, correlating to each time that an application runs. (PE)

Generation ID. A Generation ID refers to the graph created by a specific application run (a generation). (PE)

Hydration. Hydration is a process where attributes are added to portraits by backfilling additional data from the original data sources.  (PE)

Identity. An entity with a collection of traits. (PE)

JAR. A JAR is a package file format typically used to aggregate many JavaScript files, associated metadata, and resources into one file for distribution. JAR files are archive files that include a Java-specific manifest file. They are built on the ZIP format and typically have a .jar file extension. Wikipedia

Matcher. In Portrait Engine matcher is a process in an application in which records that share a blocking key are assessed to see if they should be matched to the same Portrait ID. (PE)

Metadata. Metadata are attributes of any named entity.

Merges. A merge is when two or more identities are combined in a subsequent graph generation. (PE)

Paired Data. This is a data-science-specific term referring to measuring the variance of one data set from another (often a truth set). (PE)

Pairwise matching. A methodology for specifying identifying data that selects candidates for identity resolution.

Passthrough attribute. (see attributes)

Personally Identifiable Data (or Directly Identifiable Data). PII data types can be matched directly to a person. Common examples include Name, Postal Address, Email Address, or Phone Number, but can vary by country based on the relevant privacy regulations and designations in each country.

Portrait. A resolved and unified view of the consolidated source and calculated attributes associated with a persistent Portrait ID. (PE)

Portrait Engine. LiveRamp’s first-party data identity resolution application and service. (PE)

Portrait ID. A portrait ID is a stable identifier keyed to a portrait during the resolution process. It is unique to each client and represents the entity that is being resolved in the application. (PE)

Projection. A projection refers to the vector mapping of attribute columns in the output file at the hydration stage of Portrait Engine.

Pseudonymous Data. Data that has been modified so that it cannot be reliably traced back to PII (Personally Identifiable Information).

Realm. In Portrait Engine a realm is used in the keyspace of a Portrait ID to identify if the identifier created is permissioned to sit next to known or pseudonymous data sets. For example, a graph generation process that creates an ID for use in Safe Haven next to online exposure logs would have to be pseudonymous, and therefore assigned a realm value “01” whereas a portrait created to be hydrated into a CRM would have a realm value of “02.”

Resolution. Resolution is the process of associating entities to a persistent identifier. (PE)

Resolution Application. A Portrait Engine workflow that generates portrait IDs and builds a graph based on data sources, resolution rules, or filters applied. (PE)

Ruleset. A ruleset contains multiple rules executed in a specific sequence.

Splits. A split is one kind of change that can occur to a Portrait ID from one generation to the next. Splits refer to instances in which the data that was combined in one ID in the previous generation is split into 2 IDs in the subsequent generation. One ID will retain the previous Portrait ID, and a new ID will be issued for the new ID. (PE)

Stamping. Stamping is the step in the graph generation process after entity resolution in which an identifier is assigned to the entity. (PE)

Taxonomy: The mapping of fields in a data source used in identity resolution. (PE)

Third-Party Data. Data that is aggregated from multiple parties, generally for monetization. Since this data has not been collected directly from a consumer, the consumer has not had an opportunity to opt-in or opt-out of the uses of this data, and the use of the data are subject to the terms that were in place at the point at which the data was collected. In the context of a customer application, third-party data are data from a source wherein the Portrait Engine client was not able to gather the consent and notice for their use cases.

Third-Party Graph. A graph built with data from two or more permissioned sources arising from different data sets.

Traits. Traits are attributes or combinations of attributes in a portrait derived from attributes of the original data source(s) that represents a distinguishing feature. (PE)

Traitsets. A traitset is a combination of traits, which are often specified in the blocking or matching phase to generate candidate pairs for matching. (PE)Truth Set Is a trusted source that is used for benchmark comparison for the accuracy of entity resolution, which is generally withheld from the build process. (PE)

Truth Set: This is a trusted source that is used for benchmark comparison for the accuracy of entity resolution, which is generally withheld from the build process.