In most organizations, teams build their own extracts, define terms their own way, and ship their own dashboards. Data products replace that with managed, reusable assets, and the concept is easiest to grasp through examples.
Seven types of data products come up most often:
- Source-aligned
- Master data
- Analytical
- Operational
- Machine learning
- AI-ready
- External
Most data products fit into one of these types, but their practical uses vary widely. For example, a data product could be one trusted customer record that sales and service teams both use, a live stock check that tells an online store whether an item can ship, one agreed revenue figure that finance and sales both report from, or a customer briefing that an AI support agent reads before it replies.
Each is built for a specific consumer and comes with the quality rules, business context, access controls, and interfaces that the consumer needs to use it with confidence.
There is no single agreed taxonomy for data products, so this guide groups them into these seven types by the job each one does. If you need a definition first, start with our complete introduction to data products. Here, we focus on what data products look like in practice.
Key takeaways
- There are seven main types of data products. Source-aligned and master data products are the foundations. The other five serve a specific kind of consumer and depend on them.
- Every type needs a named owner, a defined consumer, and its business context and governance attached to the product.
- Dashboards, apps, and APIs are delivery formats. The consumer decides which to use, and one product can be delivered through several.
- Master data products are a common first build, because most other types depend on one trusted version of core records such as customers and products.
What are the main types of data products?
Two types are foundations that other products build on. The other five each serve a particular kind of consumer. Use the list as a planning lens, because real products often fit more than one type. A live stock check, for example, is operational in purpose, built on master data, and delivered through an API.
Dashboards, apps, and marketplaces are missing from the list on purpose. They describe how a product is delivered or found, and the same product can use several. One analytical product might reach people through a dashboard and reach a model through an API.
| Type | Example | Main consumer | Typical delivery |
|---|---|---|---|
|
Source-aligned |
A live copy of every order from the order system, with a named owner and a documented format | Data teams and other data products |
Live feed, tables, API |
|
Master data |
One customer record that merges the same person from the CRM, billing, support, and loyalty systems | Every team and system that needs customer data |
API, SQL, shared to other platforms |
|
Analytical |
One agreed definition of revenue by region, calculated the same way for finance, sales, and leadership | Finance, sales, and leadership |
Dashboards, SQL |
|
Operational |
A live stock check an online store calls at checkout to see whether an item can be sold and shipped | Applications and workflows |
API, live feed |
|
Machine learning |
A shared set of customer details (tenure, support history, usage) that several models use to predict who might leave | Data scientists and models |
Tables, API |
|
AI-ready |
A customer briefing an AI support agent reads before replying: order history, account status, open complaints, and what it’s allowed to do | AI agents and copilots |
API, agent connection (MCP) |
|
External |
A product catalog that distributors pull straight into their own systems: specifications, compatibility, and availability | Customers, partners, and developers |
Public API, data sharing |
What does each type of data product look like in practice?
Each profile below follows the same format: what it is, an example, who uses it, practical applications, benefits, what context and controls have to travel with it, and what to watch out for.
1. Source-aligned data products: governed access to data at its origin
A source-aligned data product makes data from the system where it’s created available for reuse, with its owner, format, freshness, and quality documented.
For example, a live copy of every order from the order management system lets other teams use the data without connecting to the source themselves. Because it carries every duplicate and inconsistency from its source, it works well as an input but poorly as the final thing a business user or AI agent consumes.
2. Master data products: shared business entities
A master data product presents a core business entity, such as a customer, product, or supplier, as one trusted, governed record that every team and system can use.
A Customer 360 record, for example, matches and merges the same person across the CRM, billing, support, and loyalty systems, so duplicates are fixed once instead of in every consuming system. Teams often build this type last, after other products have already decided what a customer is, and correcting that afterward means changing every consumer.
3. Analytical data products: metrics an organization can agree on
An analytical data product packages governed data as metrics, forecasts, scores, or reports for a specific decision, with documented calculation logic and a named owner.
One agreed definition of net revenue by region, for example, means finance, sales, and leadership stop seeing different figures. A dashboard only counts as this type when it’s managed like a product, and even a well-defined metric will be wrong if the customer records beneath it haven’t been matched and merged.
4. Operational data products: data that supports a decision as it happens
An operational data product supplies current data to an application or workflow at the moment a decision has to be made, such as a live stock check an online store calls at checkout.
It returns the reason an item is blocked along with the quantity, so every consuming system acts on the same status. Operational products carry uptime and response-time requirements that analytical ones don’t, so someone has to own availability.
5. Machine learning data products: reusable inputs for models
A machine learning data product supplies the inputs models learn from in a reusable form, with consistent definitions, quality checks, and a record of where the data came from.
A shared set of customer details used by several retention models, for example, stops teams from calculating the same inputs in slightly different ways. Because training data and live data can diverge over time, versioning the inputs helps teams find what changed when a model’s behavior shifts.
6. AI-ready data products: context for agents and copilots
An AI-ready data product delivers data together with its business context, permissions, and usage guidance, so an AI agent can use it correctly without asking a person.
A customer briefing for a support agent, for example, includes recent orders, open service issues, and the consent restrictions that limit what the agent can do. This type sits on top of master data products, because an agent working from unmerged records or missing context will infer something and act on it.
7. External data products: trusted data for customers, partners, and developers
An external data product makes governed data available outside the organization under defined terms of access and use, such as a product catalog that distributors pull straight into their own systems.
Because outside consumers can’t ask an internal owner what a field means, definitions ship with the product, and changes need advance notice and versioning. Legal and privacy review should happen during design.
How do the different types of data products connect?
Most organizations use several types together, and the types depend on each other in a rough order:
- Source-aligned products supply traceable inputs.
- Master data products create one trusted version of the core entities, such as customers and products.
- Analytical, machine learning, and AI-ready products then build on those entities.
- Operational and external products put them in front of applications and partners.
That dependency is why one definition of “customer” has to reach every type. When each product defines it separately, semantic drift follows. The analytical metric, the operational API, and the AI agent each give a different answer to the same question. Defining the customer once and letting every product use that definition avoids that.
Which data product should you build first?
Four main questions can help you decide which data product to build first:
- Do several teams keep their own version of the same customer, product, or supplier? Start with a master data product. Most other types depend on it.
- Is a decision waiting on data at the moment it’s made? Start with an operational product.
- Do teams disagree about which number is right? Start with an analytical product and one approved metric set.
- Is an AI pilot giving confident but wrong answers? Check the master data beneath it, then build an AI-ready product on top.
Whichever you choose, name one consumer and one owner before you build. Without them, no one can say whether the product works.
How Semarchy delivers Intelligent Data Products
Most of the types in this guide depend on the same starting point: trusted, well-defined core business data, such as customers, products, and suppliers. The Semarchy Data Platform is built around that.
Starting from MDM as the trusted data foundation, teams design Intelligent Data Products through semantic modeling, so business context, rules, and governance are part of each product.
Each product is designed once, versioned in Git, and promoted through CI/CD pipelines. It then serves analytical, operational, and AI consumers with the same context and controls applied.
Ready to see it in practice? Explore our demos.
FAQs
What are examples of data products?
Common examples include a single customer record merged from several systems, one agreed revenue figure, a live stock check for an online store, a shared set of customer details for churn models, a customer briefing for an AI agent, and a product catalog feed for distributors. Each has a named owner and a defined consumer.
How many types of data products are there?
There is no agreed number. This guide uses seven common types: source-aligned, master data, analytical, operational, machine learning, AI-ready, and external. Other frameworks use fewer or different categories, and real products often fit more than one type.
Is a dashboard a data product?
It can be. A dashboard is an analytical data product when it is managed like one, with a named owner, approved metrics, documented calculation logic, and defined quality expectations. A dashboard without those is a report built on whatever data was available.
What is the difference between a master data product and an AI-ready data product?
A master data product provides one trusted, governed record of a business entity, such as a customer. An AI-ready data product delivers data together with its business context, permissions, and usage guidance so an agent can use it correctly. The AI-ready product usually builds on the master data product.
What’s the difference between a machine learning and an AI-ready data product?
A machine learning data product supplies the inputs models learn from, such as customer details and past outcomes. It’s used by data scientists and ML engineers, and it needs consistent definitions and versioning so models can be trained, retrained, and explained reliably.
An AI-ready data product is used by agents and copilots when they look up company data before answering or acting. It delivers data together with its business context, permissions, and usage guidance, so an agent can use it correctly without asking a person.
Which type of data product should you build first?
Start with the type that matches your most repeated problem. A master data product is a common first choice, because analytical, operational, and AI-ready products all depend on trusted core records. Whatever the type, name one consumer and one owner first.
Share this post
Featured Resources
L’état du Data Management en 2026





















































