What Are Data Products, Anyway?

There is a growing gap between data producers and data consumers in many organizations. On one side, data producers, bombarded with a fast, endless flow of data, are worrying about ingestion pipelines, security, and monitoring quality. At the same time, data consumers are challenged with how to discover data, explore it, and apply it to their specific line of business needs. This disconnect between producers and consumers can result in lack of adoption, higher costs, and longer times to outcome.

Closing this gap starts at the very beginning of your data management processes. Structured and unstructured enterprise data is ingested, transformed, and governed, and needs to be curated into a domain-centric asset that unlocks value. With a data product, you get more data ownership, accountability, and trust to your enterprise data — and a way to bridge that gap.

Defining a Data Product

A data product is an entity that allows you to gather and package trusted datasets related to the same domain or any grouping that makes sense for a specific business use case. More specifically, they are curated collections of productized datasets and business-approved metadata and domain logic designed to solve domain-specific business outcomes. Data products also bring a product management approach to managing data and are the backbone of powerful, AI-driven data apps including recommendation engines, predictive models, dashboards, and APIs, and are used across industries to extract value from data.

Data products are highly trusted, reusable, and consumable data assets. Their purpose is to bring ownership, accountability, and trust to data. They bridge the gap between data producers and data consumers.

What Does a Data Product Include?

Data professionals have always treated data as a byproduct of code and pay close attention to versioning of code. But with data products, code is not looked at as something separate — it is bundled in with the data product. It has business semantics and domain knowledge encoded into it, making it easy for data consumers to understand the intent of the data assets, and guarantee that the data quality is suited for the use case at hand.

To ensure the data product is TRUSTED, it has data quality rules, privacy policies, and data contracts with service level objectives (SLOs) clearly defined. The data product also provides standardized interfaces for access, such as SQL, REST API, event, or a vector search interface.

Just like any other product, a data product is built iteratively through various versions and taking consumer or customer feedback into consideration. This means that no one will start with a data product which has all the qualities of a more mature version. It may start with a simple dataset (which could be a few tables), some metadata, a few quality rules, and be exposed via a SQL interface. But it will be documented. It will have an owner. It will be published in a catalog. It can be trusted. It will serve a business purpose. It will have a service level objective.

The Four Types of Data Products

Just as there are different kinds of data, there are also a variety of data products. Most fall into one of the four categories explained below:

1. Source aligned: these may be, for example, data products created around your SAP application or data products around Workday.

2. Core/foundational: these data products are more enterprise focused. For example, in healthcare it could be a member data product; or in a bank, it could be a customer interaction data product.

3. Consumption aligned: this type of data product is within domains such as marketing or finance. Typically, they are more aligned to specific use cases.

4. Derived/aggregated: this variety of data product leverages other, previously created data products.

The first two types of data products above are likely to be created by a centralized team and involve complex data acquisition, modeling, and transformation. The remaining two are going to be created more often by line-of-business- or domain-focused teams.

How Data Products Are Used

Data consumers interact with data products via self-service but controlled access to data through visualization, APIs, SQL, and other secure options. Users pick existing data products from a data marketplace that match their use cases, such as a sales order data product, opportunity data product, accounts receivables data product, employee data product, and customer data product. The data product itself may be used for reporting, analytics, or data integration.

Data products are most often stored and found by users in a data product catalog. Such catalogs allow for data product management and also serve as a data product marketplace (where data consumers can search, find, and consume data products).

But What Do They DO? Examples of Data Products

Here are examples of data products in action:

• Decision Support. GPS navigation applications serve as decision-support data products. They provide real-time guidance to users, helping them make informed decisions about routes and directions.

• Algorithms. One of the most well-known examples are recommendation engines used by companies like Netflix and Amazon to suggest products or content based on user behavior and preferences.

• Automated Decision Making. A self-driving car exemplifies automated decision making. It relies on complex algorithms to make real-time decisions without external user intervention.

Key Elements of Data Products:

Data products are domain centric, modular and reusable, and are built by a designated owner using product management principles. A data products catalog organizes all data products within your organization. Here we describe each aspect in more detail:

• Domain Centric. The domain model enhances data understanding and adds business context. It does this by housing the business logic of transformations, analytics calculations, metrics, and machine learning. It acts as a semantic layer, exposing the business-friendly domain data and insights, while abstracting the technical details.

• Modular And Reusable. They are built once and then reused multiple times in various use cases, so new ones can be built using existing ones.

• Ownership. Each one is owned and managed by a domain-aligned data team that is responsible for its success in delivering value, satisfying and growing data users, and maintaining its lifecycle.

• Product Management Principles. Like all products, improvements are made iteratively and are informed by user experience. Every data product goes through versions and enhancements based on customer feedback. There is also a definitive lifecycle that must be maintained.

Business Impact of Data Products

Ultimately, data products help you bridge the growing gap between data producers and data consumers to shorten the time to value for a business’s use cases. Early studies have shown that data products can offer 90% faster use case delivery and a 30% reduction in total cost of operations (TCO).

The Verdict on Data Products

Data-driven organizations require a unique and robust approach to unleash their data's full potential. The data product offers a unique, product-centric approach to streamline end-to-end data management by providing high-quality, curated, and readily accessible, domain-centric trusted data. This aligns data and analytics stakeholders and empowers them to unlock the full value of data and accelerate business outcomes, while enhancing data trust and governance.

Learn more about data products here.