MarTech architecture glossary

A CDP brings together customer data from multiple sources, resolves identities, and maintains a persistent profile available for audiences and activation.

A CRM maintains the operational record of known customers, accounts, opportunities, commercial interactions, and service activity.

A DMP organizes mainly pseudonymous signals and audiences for advertising buying and activation. Its historical model depends on identifiers with a limited lifetime.

An Identity Graph links observed identifiers to represent which fragments may belong to the same person or entity. It resolves evidence; it does not invent relationships.

Identity stitching is joining the identifiers one person leaves across channels, such as a cookie, an email, and a customer number, so the system treats them as a single profile.

AEP is the Adobe foundation for customer data. It provides ingestion, a shared data model, identity resolution, a unified profile, and query. The applications run on top of it: Real-Time CDP, Journey Optimizer, and Customer Journey Analytics.

Journey Optimizer is the Adobe application on AEP that orchestrates journeys and sends messages over email, push, SMS, and custom channels. It reads from the unified profile and can write back to it.

Customer Journey Analytics is the Adobe analytics application on AEP. It queries data lake datasets through a connection and lets you join channels using whichever identity you choose as the person ID.

The Data Lake is the queryable history in AEP, where datasets live and what Query Service and Data Distiller work against. The Profile Store, by contrast, is the live view used by audiences, journeys, and activation.

Data Prep transforms source fields into XDM during ingestion. It can derive values, normalize structures, and apply functions before writing to the dataset.

A dataset is the container where AEP stores records conforming to an XDM schema. It can receive batch or streaming data and feed downstream services.

A namespace declares what an identifier means, such as a CRM ID, email address, or device. The value only has identity meaning within that context.

Profile combines enabled fragments and identity relationships to provide an operating customer view for segmentation, personalization, and activation.

Query Service runs SQL against the datasets in the AEP data lake. It is used to explore data, validate ingestion, build derived tables, and feed reporting.

Real-Time CDP is the Adobe application on AEP used to build and evaluate audiences, enforce data usage policies, and activate them to destinations.

The Web SDK is the Adobe client library, known as alloy, that sends events to Edge Network in a single call. From there they are distributed to Analytics, Target, Real-Time CDP, and the rest.

An XDM schema defines the structure and meaning of records entering AEP. It combines a class, field groups, and declared identities.

A composable CDP distributes collection, modeling, identity, audiences, and activation across connected components, usually with the warehouse as the source of truth.

ETL transforms data before loading it into the destination. ELT loads first and uses the destination platform to transform it afterward.

An iPaaS provides managed tools for connecting applications, transforming messages, orchestrating flows, and monitoring reusable integrations.

Reverse ETL synchronizes models, attributes, or audiences computed in the warehouse into operational tools such as CRMs, engagement platforms, or advertising destinations.

A warehouse-native architecture models profiles and audiences on the warehouse and exposes them to activation tools without maintaining another central copy as the source of truth.

Zero-copy allows a platform to query or use data that remains in the warehouse without maintaining a complete replica in its own store.

Activation makes an audience, attribute, or decision available to the system that must act, under the applicable identity, consent, and usage rules.

A CEP coordinates customer interaction through audiences, journeys, decisions, and channels. Its center is acting on the customer, not keeping all history as the source of truth.

Decisioning evaluates eligibility, priority, and constraints to choose content, an offer, or an action in context. It can operate in batch or during a request.

A journey coordinates steps, waits, conditions, and actions from an entry. It maintains operating state about the flow, not the complete customer truth.

A MAP automates campaigns, lead nurturing, scoring, and sales handoffs, usually with email and B2B processes at its core.

Next Best Action is a decisioning pattern that compares eligible actions and selects the most appropriate one for a person and context.

CIAM manages registration, authentication, account recovery, and account security for customers using digital channels.

A CMP captures and stores consent choices and exposes signals so sites, tag managers, platforms, and destinations can enforce them.

Consent Mode is the Google mechanism that lets its tags adjust behavior based on what the user has consented to. Version 2 adds two advertising signals to the analytics and storage ones.

A data clean room allows several parties to compare or analyze data under controls that limit access to individual records and sensitive outputs.

A Conversions API sends events and conversions from infrastructure controlled by the organization into an advertising platform.

The data layer is the structure a page publishes so measurement reads from it instead of from the HTML. Field names, types, and when each value becomes available.

GA4 is the current version of Google Analytics. Everything is an event with parameters, instead of sessions with pageviews and custom variables as in Universal Analytics.

Google Tag Manager is a tag manager with two forms: a web container that runs in the browser and a server container that receives events and distributes them from your own infrastructure.

MMM uses statistical models on aggregated data to estimate how spend, channels, and external factors contribute to an outcome.

MTA distributes credit across several observed touchpoints in journeys that end in a conversion.

Server-side means sending events to your own server and letting that server, not the browser, distribute them to each destination. The browser stops talking to every tool and talks to one.

Signal loss describes the reduction in observable data caused by privacy controls, browser restrictions, closed ecosystems, and user choices.

A TMS is the layer that decides what loads on the page, when, and with which data. It centralizes tags, trigger rules, and extensions so the site code does not change for every update.