This glossary defines Treasure AI products, platform concepts, and technical terms used in the documentation. Terms follow the Product Overview and the Customer Intelligence Loop: Collect → Unify → Understand → Decide → Engage.
Jump to: A · B · C · D · E · F · G · H · I · J · M · N · O · P · R · S · T · U · W
An activation sends a segment or journey audience to a destination — a messaging channel, advertising platform, or other external system — so those profiles can be reached.
The Activation pillar of the Agentic Experience Platform closes the Customer Intelligence Loop (Decide / Engage). It contains Engage Studio and Personalization Studio. In the portal navigation this group is labeled AI Suites.
Account and user administration outside the three product pillars: Control Panel (users, permissions, security policies, SSO, audit logs) and My Profile.
Treasure AI's platform: a unified control plane where AI agents run the Customer Intelligence Loop continuously, while humans set direction, define guardrails, and bring the judgment that makes customer experiences feel human. The platform has three pillars — AI, Activation, and Context (Treasure Data CDP).
The AI pillar is the command center for agents and customer data — Web, Desktop, Mobile, CLI, and Voice. It contains Treasure AI Studio, Treasure Code, and Treasure AI Voice. Treasure AI Labs is the R&D surface next to this pillar.
An AI agent is software that reasons over unified customer data and takes actions inside human-defined guardrails — querying data, building segments, drafting campaigns, or generating reports. Treasure AI Studio orchestrates purpose-built agents; AI Agent Foundry is where custom agents are built and governed.
AI Agent Foundry is the Treasure Data CDP component where AI agents are built, governed, and grounded in the unified customer profile (the Diamond Record).
The navigation label for the Activation pillar: Engage Studio and Personalization Studio. The directory slug remains marketing-cloud/.
Purpose-built, domain-specialized agents orchestrated by Treasure AI Studio — for example, super agents for marketers or for data analysts. Not a product name.
A standard for SQL specified by the American National Standards Institute (ANSI).
A stored credential that lets Treasure AI connect to an external system (database, cloud storage, SaaS, or marketing tool) for import and export. Multiple integrations can reuse one authentication. Authentications are managed in Integration Hub.
In database systems, atomicity is one of the ACID (Atomicity, Consistency, Isolation, Durability) transaction properties. An atomic transaction either fully occurs or does not occur at all, so another client never sees a partial update.
In Treasure AI, Append, Replace, Truncate, and Update are atomic because incoming data is written to a temporary table before the destination table is modified.
Audience Studio builds and activates segments on unified customer profiles (Understand / Decide in the Customer Intelligence Loop). It is a Complete CDP component. See Audience Studio.
Automated Machine Learning — notebooks and operators that automate training, prediction, and related steps so users who are not ML specialists can build models. AutoML is part of the Machine Learning component of Treasure Data CDP.
The AutoML base package provides a generic AutoGluon notebook that you configure for a use case, plus general ML training, prediction, and dataset-loading operators.
A setting that lets one user train an AutoML model and another user run that model on their data to make predictions.
A schema-based serialization format commonly used with Hadoop. Serialized data is binary and is accompanied by the schema needed to deserialize it.
A collection of profiles selected by attribute and behavior rules, evaluated on a schedule, and used for activation and reporting.
The Integration Hub catalog of connectors for importing and exporting customer data. See integrations.
California Consumer Privacy Act — rules that govern how the personal information of California consumers is collected and processed.
Complete CDP is the fully managed Treasure Data CDP deployment: collect, unify, analyze, and activate customer data without requiring an existing data warehouse. The other mode is Composable CDP. See Complete CDP.
Another name for Composable CDP — the warehouse-first, zero-copy Audience Studio deployment on Snowflake, Databricks, or BigQuery.
Composable CDP is the warehouse-first Treasure Data CDP deployment: a zero-copy intelligence layer on Snowflake, Databricks, or BigQuery. Customer data stays in the warehouse. Also called Composable Audience Studio (CAS). See Composable CDP.
Control Panel is where administrators manage the Treasure AI account: users, permissions, security policies, SSO, and audit logs. See Control Panel.
A website banner and related JavaScript SDK options for collecting end-user consent before setting cookies.
A time-based scheduler that runs jobs at fixed times, dates, or intervals.
The loop every Treasure AI product exists to run, feed, or accelerate: Collect → Unify → Understand → Decide → Engage, then feed results back to Collect. Context (Treasure Data CDP) builds the loop, Activation closes it, and AI runs it continuously.
A directed acyclic graph — a representation of tasks and their order. Vertices are tasks; directed edges are dependencies. Acyclic means there are no loops.
Merging first-party and third-party data into a dataset to add attributes.
The structure of your data: how tables, columns, and attributes describe customers.
A customer-dedicated PostgreSQL-compatible data mart for interactive analytics, reporting, and operational access to data in Treasure Data CDP. Data Tank 2.0 is the current offering; Data Tank 1.0 is obsolete. See About Data Tank 2.0.
Data Workbench is the Complete CDP component for databases, tables, Trino and Hive queries, and workflows that prepare data (Collect / Understand). See Data Workbench.
A named collection of tables in Data Workbench, designed to be queried, managed, and updated.
A journey step that splits profiles onto different branches based on attributes or activity.
The Diamond Record is the unified, trusted, activation-ready customer profile produced by the Unify stage of the Customer Intelligence Loop. AI agents reason over this profile; it is the grounding source that keeps answers tied to customer data.
The open-source workflow engine that Treasure Workflow is based on. Most, but not all, Digdag operators work in Treasure Workflow.
Engage Studio is the native customer engagement platform built on Treasure Data CDP: design journeys, send Email, Mobile Push, SMS, LINE, and related channels, and optimize campaigns with AI. See Engage Studio.
The condition a profile must meet to enter a journey stage.
Extract, transform, and load — moving data from multiple sources into a warehouse or other unified store.
A tracked customer action or system occurrence used for analytics, real-time attributes, and activations.
A container used to organize objects in the console (segments, journeys, workflows, and similar).
The workspace in AI Agent Foundry where you configure, test, and iterate on agents before deployment.
A column (or set of columns) that enforces referential integrity by pointing at a key in another table.
A simple staged workflow that profiles move through. Each stage can be configured separately.
General Data Protection Regulation — rules governing privacy and processing of personal data for people in the European Union.
An installable Ruby package. RubyGems is the package manager.
A database of geographic and related information about IP addresses. Treasure AI geolocation features are based on MaxMind.
An AutoML base-package notebook that scores data with a model built by Gluon Train.
An AutoML base-package notebook that builds a prediction model with AutoGluon, automating data processing, feature engineering, model selection, ensembling, and hyperparameter tuning.
Google Tag Manager.
An SQL query engine for large-scale batch processing. Hive queries can complete even when individual tasks fail. Use Hive when a Trino query hits memory limits or you need Hive-only syntax.
ID Unification links fragmented identifiers (cookies, emails, device IDs, and others) into one canonical customer profile (Unify in the Customer Intelligence Loop). See ID Unification.
Integration Hub is the Complete CDP component that provides 200+ pre-built connectors for getting data in and out (Collect / Engage). See Integration Hub and the Integrations catalog.
Connectors that import data into Treasure AI or export it to external systems. Batch and streaming integrations are configured in Integration Hub; each connector has a catalog page under /int/.
A driver that lets applications connect to Treasure AI databases and run SQL.
Job Management is the Complete CDP component for monitoring platform jobs. See Job Management.
A shared dimension used to join data sources.
A multi-stage orchestration that moves profiles through stages (for example awareness, consideration, purchase, retention) based on attributes and behavior. Journeys can run on a batch schedule or in real time. See Journey Orchestration.
Journey Orchestration is the Complete CDP component for state-based omnichannel journeys that react to individual behavior (Decide / Engage). See Journey Orchestration.
The Complete CDP component for predictive scoring: AutoML (no-code notebooks) and Precision ML (Python). Understand / Decide in the Customer Intelligence Loop. See Machine Learning.
An API key with full access for the operations the user is permitted to perform, including read and write. Keep Master keys private; do not embed them in public pages or client-side code.
All sequential-enabled activations that share the same authentication. Sequential-enabled activations do not run in parallel with their matched activations.
A Java tool that manages versions and dependencies during a build.
A journey step that brings decision-point branches back into a single path after activations on those branches complete.
A method of marketing measurement that assigns value to each touchpoint that led to a conversion. Treasure AI provides an MTA AutoML notebook for this analysis.
The signed-in user's personal settings. See My Profile.
A user-defined workflow triggered by an activation. It processes the activation output before export. The activation succeeds only when both the activation and the nested workflow succeed.
A strategy for choosing the marketing action most likely to move a customer toward a conversion. Treasure AI provides an NBA AutoML notebook for this analysis.
Open Catalog Access publishes CDP tables in Apache Iceberg format so Snowflake, Databricks, or BigQuery can query them in place — read-only, zero-copy (Understand). See Open Catalog Access.
A unified data model (also called Master Segment) that joins a master table to attribute and behavior tables. Parent segments are the foundation for child segments in Audience Studio and refresh on a schedule.
Personalization Studio is AI-native web personalization: different images, text, and CTAs for different audiences in real time, powered by the CDP. See Personalization Studio.
Personally Identifiable Information.
The package installer for Python.
An open-source relational database. Data Tank runs on PostgreSQL.
Using historical and activity data with a model to predict a future behavior, such as conversion or churn.
The former name of the Trino distributed SQL query engine. Current Treasure AI documentation and the query UI use Trino. Legacy pages, JDBC/ODBC driver names, and older release notes may still say Presto.
The record of a person or account — attributes, behaviors, and identifiers — used to segment and activate audiences. The unified, trusted form of this record is the Diamond Record.
Real-Time is the Complete CDP component for real-time ingestion, attributes, and triggers, so the Customer Intelligence Loop can close in minutes instead of a batch cycle. See Real-Time.
An HTTP API that follows REST constraints. Treasure AI publishes several REST APIs; see API Documents.
An operation that sends query results from Treasure AI to an external destination (database, cloud storage, warehouse, or BI tool). Result exports can run once or on a schedule.
The package manager for Ruby. A gem is an installable package of Ruby code.
A named collection of profiles that share attributes or behaviors. Child segments are built on a parent segment in Audience Studio (or Composable Audience Studio) and can be activated.
An activation with a Nested Workflow that does not run in parallel with other sequential-enabled activations that share the same authentication (its matched activations).
SHapley Additive exPlanations. Treasure AI provides a multi-touch attribution AutoML notebook that combines LSTM deep learning with Shapley values to score touchpoints in a user journey stored in Treasure Data CDP.
A configuration that defines where and how to import data into Treasure AI — connection, format, schedule, and transform rules. Sources are managed in Integration Hub and support batch and streaming ingestion.
Structured Query Language. Treasure AI query engines are Trino and Hive.
A step in a journey that describes where a profile is on the path to a goal. After you create stages, you can analyze them and activate specific stages.
A profile that has not completed a journey stage goal after a configured amount of time. Stale profiles can be captured as a segment or attribute for redirects, other segments, or reporting.
Labels on columns that classify the kind of data they hold (for example PII, finance, or sensitive) so access can be restricted.
The legacy td command-line client. New CLI work uses Treasure Code (tdx). Toolbelt pages remain for existing td workflows.
The command name of Treasure Code — for example tdx query, tdx wf, tdx sg.
Fitting a model to historical, time-stamped data to predict future values. Treasure AI provides an AutoML notebook that trains time-series models and forecasts.
The company and platform brand. Use Treasure AI for the company and the Agentic Experience Platform. The CDP product name remains Treasure Data CDP.
The R&D arm of Treasure AI, where experimental products reach early adopters before graduating to Treasure AI Studio. See Treasure AI Labs.
Treasure AI Studio is the chat-based AI application for Web, Desktop, and Mobile: query data, build segments, orchestrate journeys, and generate reports from a single conversation. Treasure Code is its CLI surface and Treasure AI Voice is its voice surface. See Treasure AI Studio.
Treasure AI Voice is enterprise voice intelligence (PLAUD AI hardware) that turns meetings and conversations into customer data. See Treasure AI Voice.
Treasure Code is the AI-native CLI for engineers (tdx), with pre-built agentic skills for query optimization, workflow syntax, and CDP data models. It integrates with Claude Code, Cursor, and similar tools. See Treasure Code.
The Customer Data Platform product — the one product name that keeps "Treasure Data". It ships as Complete CDP or Composable CDP. Context pillar of the Agentic Experience Platform. See Customer Data Platform.
Treasure Insights is the Complete CDP reporting component: dashboards on customer data (Understand). See Treasure Insights.
A task engine that orchestrates jobs on Treasure AI and other clouds. Treasure Workflow is based on Digdag. See Workflow.
The distributed SQL query engine used in Data Workbench for interactive and medium-scale queries (formerly named Presto). Trino queries do not tolerate worker failures; use Hive when a query needs Hive-only features or hits Trino memory limits.
User-defined function — custom SQL logic registered with a query engine.
Seconds elapsed since 1970-01-01 00:00:00 UTC. Example: 1672338261.
A visualization of how much Treasure AI resource capacity is being used.
A 128-bit Universal Unique Identifier used to label an object.
A journey step that waits a configured time for a profile to complete a goal before the profile continues.
Treasure Workflow is a multi-cloud task engine that orchestrates tasks on Treasure AI and on clouds such as AWS and Google Cloud. It is based on Digdag; most, but not all, Digdag operators work in Treasure Workflow.
An API key that can import data into databases the user can write to, but cannot read or delete data. Use Write-only keys for third parties and client-side ingestion so a Master key is not exposed.