Data services support the emergence of data offices and data-centric architectures to allow organizations to develop opportunities for improved data management and usage
Data Service is a part of a data-centric architecture, and the basis of an enterprise data management function. A good data-centric architecture has many benefits and organizations should focus on clarifying the reasons, opportunities and strengths of this architectural style.
What is a data service?
The emergence of data offices is combined with the development of data-centric architectures and data services. But, what is meant by “data service”?
There are very few definitions of “data service” in the literature, despite a widespread use of this expression. A data service is an interface to a business process that receives or provides data attributes, usually via a web application.
This definition focuses on the role of a process (supplier or consumer) in the exchange of data rather than on the data management capabilities required. “Data as a service” can be considered as a variation of this definition in that it assumes that data attributes can be provided to a client on demand.
A data service may involve data quality management functions, master data, metadata, data architecture, data modeling, etc. Also, it may include resources, other than data management functions, such as processes implemented, applications, infrastructure, personnel, and other activities. Therefore, a data service should be considered to represent all shared and purpose-oriented data management capabilities exposed to clients. To illustrate, consider a company which implements a CRM project and a BI project that use both customer data matching and de-duplication functions. In a first situation, the implementation of these functions can be carried out in silos, as suggested by the left half of the diagram below, each project obeying essentially its own capabilities. In an alternative configuration, we can imagine pooling them for both projects, as the right half of this diagram suggests. By doing so, we build reusable capabilities for both projects, as well as for other efforts.
Figure 1: Data Services vs. Non-Data Services Implementations
Let’s take another illustration from the pooling of reference and master data management functions:
Within a company, customer data can serve the needs of several business-functions, including sales, management, marketing, after-sales, accounting, and so on: These are shared data. Although shared, these data can be managed in silos, each business-function implementing its own management capabilities.
The alternative approach would be to pool the management capabilities of these data, which are shared by nature, and to offer master data services: this is the vocation of the Master Data Management, which can be considered as one of the first forms of a data-centric architecture.
What is a data-centric architecture?
The traditional service-oriented architecture (SOA) (or left-hand side of the diagram below) is an architectural form called process-centric: It integrates data and business processes by creating reusable components, known as business services.
In the previous example, the CRM project implemented matching and deduplication for its own account. This may be considered as an implementation carried out as part of the customer data analytics process.
Combining data with business process does not allow pooling data management capabilities and reusing them for other projects. This prevents the knowledge sharing and the emergence of a common enterprise language. Finally, combining data and processes does not break up the data management silos and does not facilitate the implementation of data governance.
Data-centric architecture is a service-oriented architecture in which data management capabilities are pooled and offered as services called data services. As shown in the right part of the diagram below, a so-called data-centric SOA architecture introduces an additional abstraction layer. Consisting of data services, this layer allows all data consumers to benefit from the same services to know, search, access and use shared data.
The value of this architectural style lies therefore in its disposition to:
- Improve knowledge and data sharing;
- Improve the consistency and reliability of data;
- Simplify and optimize access to data;
- Reduce risks and redundancies (code, functions, etc.) in the system landscape;
- Facilitate the establishment of data governance;
Figure 2: Data-centric SOA architecture
What are the reasons, opportunities and benefits of these changes?
Why create shared capabilities? Why create a data office?
It is essential to note that data sharing among front office and middle office or between back office and middle office is supported by the enterprise architecture classic scheme (fine lines on the diagram below). This architectural style is, however, unsuitable or inefficient for organizing the data sharing between:
- Front and back office layers;
- Internal layers (front office, middle office or back office) and external actors (client, authority (regulatory, fiscal, etc.), etc.).
This requires a data-centric enterprise architecture (see diagram below).
Figure 3: Data-centric Enterprise Architecture
The desire to create shared capabilities in terms of data management can have different motivations, which depend on the company profile, and even on the challenges it sets itself:
For example, having a real-time view of the available stock is a major challenge for many online retailers. But, the answer to this requires strengthening front office integration, where is located the website, with back office, where is located the inventory management system.
Adapting to customers’ wishes and behaviors in the choice of contact channels or seeking a complete knowledge of a customer are other examples requiring integration between the internal client layers.
There are many expected benefits to shared data services:
- Extend or transform the company’s service offering;
- Respond to new behaviors, consumption patterns and uses;
- Meet (strong) requirements for availability and quality of service;
- Improve the quality of information, to make reliable or speed up decision-making, etc.;
- Reduce misunderstanding internally and externally (supervisory authority, customer, etc.);
- Improve the performance of processes;
- Make more reliable internal and external control processes;
- Make more reliable the justification (accounting, etc.);
- Reduce information-processing the cost;
- Develop turnover.
A data-centric architecture is a means of building shared capabilities for data management, enabling many opportunities for organizations. Some of these capabilities include:
- Data management services catalog, to standardize the offering made to customers;
- Common language, to share knowledge and improve interoperability;
- Data management standards, to facilitate reuse, increase productivity, reduce lead times, reduce costs, standardize quality and safety;
Data governance, to set the common framework for sharing data management capabilities and monitor its implementation.