If your organization doesn’t already have an enterprise information architecture in place—it should, and if you do have one, it should be based on a semantic model. In this article, we’ll explain what an “enterprise information architecture” is and how it can support your enterprise with decision intelligence, knowledge democratization and enterprise-wide optimization.
Why your enterprise information architecture needs a semantic model
Imagine what your organization could accomplish with a complete and comprehensive digital representation of its entire IT structure. This would include mapping information such as the applications and physical systems used for the business, detailed descriptions of existing processes, as well as roles, departments and the individuals who fill them.
An “enterprise information architecture” (EIA)—not to be confused with an ‘information architecture’ or even ‘enterprise architecture’ (though it does embrace elements of both)—does precisely that. With an enterprise information architecture based on a semantic model, you can move away from documents and diagrams to having a detailed digital twin of your organization’s IT landscape that makes information relevant to data, IT and technology management and planning, available across systems, processes and departments, and made interpretable for relevant stakeholders and even machines.
In this article, we’ll explain what an “enterprise information architecture” is and how it functions. We will also discuss existing data challenges that an EIA can address and explain how a semantic knowledge model is crucial to building a robust and seamless EIA.
Table of contents
-
What is the value of having an enterprise information architecture?
-
Why your enterprise information architecture needs a semantic knowledge model
-
Building a semantics-driven enterprise information architecture
What is an enterprise information architecture?
Not everyone defines ‘enterprise information architecture’ in exactly the same way. For example, at Gartner, an enterprise information architecture is considered “part of the enterprise architecture process that describes — through a set of requirements, principles and models — the current state, future state and guidance necessary to flexibly share and exchange information assets to achieve effective enterprise change.”
For us at metaphacts, we are expanding Gartner’s definition with further specificity. To us, an enterprise information architecture includes a digital asset at its core called the semantic layer (or model), which holds all required metadata from all involved enterprise systems and stakeholders to facilitate communication, planning and coordination. It enables you to model, assess, structure, analyze, organize, manage and visualize an organization’s data and information assets across various systems, technologies, departments, processes and stakeholders – all in a human- and machine-interpretable format.
We believe that at the heart of an EIA should be a semantic knowledge model. An EIA based on a semantic model transforms this mapping of enterprise metadata of associated IT systems and processes into a powerful tool that effectively helps you make informed business decisions and drive change across the enterprise, all the way from the C-suite level to the IT team. Capturing this information in an EIA with a semantic model can support the optimization, migration and consolidation of legacy systems, as well as the introduction of new ones. Moreover, it helps you with efficient decision-making, information sharing and process optimization within an organization, while ensuring all is aligned with privacy and security requirements. Traditional methods of mapping enterprise information architecture rely on inflexible diagrams and documents and often fail to include this metadata and crucial context that is valuable for decision-making.
Our definition of enterprise information architecture is concerned with context and hidden connections revolving around enterprise systems and processes, such as when a system was deployed or where the server is located, which business objects it supports, etc. Rather than simply capturing what these systems are, our definition of an EIA also captures what they are about, how they connect to the enterprise and its other systems, why they are used and who is responsible for them.
What is a semantic knowledge model?
A semantic knowledge model explicitly defines domain-relevant objects and concepts, and the relations between them. Semantic models enrich data with context and meaning, as metadata, that is both human-understandable and machine-interpretable.
All organizations have their own terminology, concepts and intricate connections between these, which are unique to their company and industry. Semantic models can capture this domain-specific knowledge — whether from structured text, unstructured text or if only available in the minds of domain experts — into a central repository. This creates a shared understanding of these concepts, terminology and their relationships, and bridges the semantic gap between departments or colleagues, thereby fostering better collaboration and decision-making.
In the context of an EIA, a semantic model explicitly describes granular details about the “business objects”, systems and processes that exist in the organization and how they work. This metadata is the nuance that may only be known by individuals who’ve been at the company for many years or documented in emails and paper files, yet can make all the difference when making key strategic decisions.
The main elements within an enterprise information architecture
Business Objects
Business Objects are the fundamental building blocks of a consistent and enterprise-wide information architecture and:
- Represent elements of the business, real or virtual (e.g., sales order, plant or engine specification).
- They are also the input, output or interim result of a business process, such as the results generated from a sales pipeline, meaning they can also be created, transferred or modified within the process.
- These Objects are named using a term that is harmonized and consistent within the enterprise (through a defined vocabulary).
- Lastly, they are represented and persisted by data within IT systems.
Physical Objects or representations
Physical Objects are the representation of a Business Object on the physical layer of the EIA. They are physical data models or schemas that represent the physical structure of data of a business object in a specific context.
Physical Objects are part of the IT landscape, and cover complete or parts of a Business Object. One Business Object can have multiple physical representations in different IT systems, databases, data warehouses, the data lake or data transport technologies.
Business processes
Business processes in this context refer to workflows or series of related tasks that are performed regularly by specific stakeholders or systems to achieve an organizational goal.
An example of a business process is invoicing. Business processes can involve several business objects—such as customers, products and contracts in our example—as well as multiple IT systems, such as a CRM system, an invoicing tool and a reporting tool.
IT systems
When we refer to IT systems, we mean applications such as CRM systems like Pipedrive or Salesforce. Meanwhile, physical objects are the actual physical servers where the IT systems are hosted.
What is the value of having an enterprise information architecture?
When you are equipped with information about your IT landscape and have captured the nuances of how your organization operates, you can then derive sophisticated insights about your company, which can be used to inform strategy, planning and coordination, communication and decision-making on a small and large scale, such as with:
-
Scaling operations—e.g., opening new facilities
-
Market expansion—e.g., entering foreign markets
-
Product diversification—e.g., introducing new products
-
Mergers & acquisitions—e.g., support with integration of systems
-
Sunsetting products—e.g., identifying & retiring products
Let’s use the example of a hypothetical company and examine its plans to expand into the Asian market, which is entirely new to them. As part of the process, one of its primary concerns will be understanding how to integrate its sales processes into this new market. This company will need to aggregate and analyze information about sales forecasts, business objectives and identify which systems are currently in use. If a new system is required for the Asian market, the company must determine which system that would be and how it could support its sales forecasts, and how it would align with existing data and processes. Additionally, they will also need to answer questions such as: What is needed to set up the expanded process? Is an additional business intelligence system, knowledge graph or data warehouse needed to integrate and analyze the data correctly, ensuring accurate data representation?
An enterprise information architecture could facilitate the transition to this new market and systems by providing detailed metadata about the associated systems and processes currently used, as the person responsible for implementing the previous system contributed their specialized knowledge into the EIA’s semantic model, making valuable information explicit that could be integral to the decision-making process. This information that is now captured in the model could then be shared with relevant stakeholders or used with AI applications.
Why your enterprise information architecture needs a semantic knowledge model
Even with an enterprise information architecture in place, your organization might still be facing significant data integration, governance and discovery challenges that hinder your ability to identify opportunities to optimize existing processes or scale the organization with efficiency and ease.
Lack of machine-interpretability
Organizations operating with an information architecture or enterprise architecture as they are understood today often use tools like UML (Unified Modeling Language) or ArchiMate diagrams to generate diagrams of their data environment. However, these diagrams are just that—diagrams, or pictures of a data architecture that includes a visual layout and is encoded in comments but are still limited in what they can offer. Machines aren’t able to interpret the information, meaning there is no way to search over them and discover information easily. Additionally, there’s no way to validate the data without human assistance, so a person must analyze the diagram and be able to accurately interpret it in order to ensure the diagram is up-to-date and captures any changes, and this process doesn’t hold up to the speed of innovation that is usually expected. At this pace, when systems aren’t in the state that they should be, the enterprise always falls behind. These diagrams can also quickly become unwieldy and cluttered with increases in data volume; it’s not as dynamic or easy to maintain and collaborate on.
Data silos & barriers to integration
Organizational data is stored in multiple applications and systems that are independent of each other (e.g., CRM, ERP), and as a result, silos form. The ‘enterprise architecture’ approach attempts to address this but isn’t able to do so sufficiently due to its inability to connect and connect to these systems. An enterprise information architecture based on a semantic model (and underlying knowledge graph) is interoperable and capable of integrating external systems and applications, as well as linking these systems to each other, therefore ensuring information is regularly updated and captures the granularity necessary for understanding how data aligns across systems.
Data quality
Challenges with data quality in this context have two meanings. First, it refers to the quality of the EA models themselves, without a semantic model at its core, these models are often outdated, incomplete and missing relevant context and information, and they can’t be quality-checked in an automated way if they are not fully machine-interpretable.
Secondly, data quality issues in this case can also refer to the quality of the data within the systems themselves. For example, there’s no easy way to validate if the data inside a particular system complies with its data models and definitions (which would potentially be captured in a semantic model inside of an EIA, if one was used). There would also be no way to check the consistent use of names, identifiers or other references for data between different systems.
Cost
An EIA can help alleviate some of the IT costs associated with conducting any large-scale enterprise project, such as data analysis or migration related to market expansion or opening new facilities. For example, one metaphacts customer’s growth, especially through acquisitions, resulted in them running a large number of ERP systems, but they realized that the information they had in their independent information architecture was so limited that whenever they wanted to migrate or consolidate systems, they needed to get external consultants involved to assess the systems in detail. These kinds of projects could require several people over an extended period, equalling roughly a six-to-seven-figure investment for each request. On top of the cost of changing or migrating systems, there is also an additional cost of lost opportunity, resulting from the inability to identify efficiencies that could save the organization time and money.
Low stakeholder engagement
The diagrams made using common enterprise architecture tools are inflexible and can’t be easily broken down into different views, making it difficult to arrange for new use cases or stakeholder views. When information is unusable and inaccessible, it inhibits collaboration and engagement among stakeholders.
What is different with a semantic layer?
Besides addressing the above-mentioned shortcomings of current approaches, the semantic layer, based on open standards (W3C Semantic Web Standards, including OWL, SHACL and SKOS), helps organizations to build a data culture with data democratization or, what we call knowledge democratization. Knowledge democratization enables everyone in the enterprise, including business users—and not just the IT teams—to gain a sufficient understanding of the data resources available in the organization, know how they can support their business processes and daily needs, and how to utilize these resources for specific information needs and decisions. This knowledge democratization is also the foundation for achieving AI democratization in the organization. It helps users to better understand results generated with AI, enabling them to more effectively apply the technology on top of the data available in the organization. It becomes the tool for everyone to communicate and understand each other when it comes to data, knowledge and AI.
Our experience is that such transformation is not just a tool topic. For seamless and sustainable implementation, teams need adequate training on related methods and concepts. It can range anywhere from 2 hours for a business user who only needs basic data and AI proficiency, to 4-5 days for power users who will actively help develop the semantic layer. For the later users, we have also seen approaches where joint modeling (e.g., via walk-in sessions) is helpful in addition to the training, to get them up to speed quickly.
The semantic layer closes the gap between existing data (governance) tooling, from data catalogs to business glossaries, and even CMDBs, eGRC solutions, data lakes or data API gateways or portals. It acts as a glue between all of these systems, seamlessly connecting business terms (e.g., from the business glossary) to relevant data sources (via the data catalogs), and to available services (via the API gateway or data portal). However, it doesn't need to be the leading interface for any of these tasks; if you want, it can simply function as a connecting element, serving relevant information via APIs into all other systems.
Nevertheless, the semantic layer requires an interface to model it, and link between the different systems and or “worlds”. While this can be supported by AI (through automated extraction and suggestions), there should always be a human in the loop to validate and ensure quality for the metadata in the semantic layer. Furthermore, the semantic layer can also act as an exploration and discovery interface for the models across systems, adapting to the language of each user (e.g., business users, IT users, etc.), and delivering relevant information with their specific context. As such, it can virtualize or federate over the metadata and even function as a conversational AI interface for users, enabling users to input queries using natural language. For example: “Given my role in procurement, and needing to select the best vendor for the given services, what data is available in our organization on previous vendors supplying similar services, and what is their track record?” The system would then respond with a list of relevant source systems and their owners, along with the type of information available in each system, and it can even provide virtual access to the actual data within those source systems.
Building a semantics-driven enterprise information architecture
Our enterprise knowledge graph platform, metaphactory can help facilitate the creation and management of your EIA. metaphactory is a FAIR data platform that empowers you with semantic knowledge modeling and knowledge discovery capabilities. It enables you to capture and organize business-specific knowledge in explicit semantic models, extract insights from your data and democratize knowledge across the enterprise.
Semantic knowledge modeling
In metaphactory, the modeling of the enterprise information architecture happens in three layers:
-
Conceptual layer: Modeling of business-relevant concepts (Business Objects) and their attributes; these Business Objects are represented and persisted by data within IT systems and that help organize, structure and manage business data that is exchanged between business processes
-
Logical layer: Capturing a Business Object’s logical definition and its relations to other business-relevant concepts
-
Physical layer: Capturing of connections to IT systems that store the physical representation of concepts
What facilitates this modeling? The technical features of metaphactory’s semantic knowledge modeling include:
-
Visual ontology editor
-
Vocabulary & taxonomy management
-
Data catalog integration
-
Collaboration, versioning & metadata curation
-
User-friendly modeling interface for collaboration with non-tech experts
-
System based on open standards to facilitate interoperability and reusability
For a step-by-step guide on creating a semantic layer for your enterprise information architecture, watch this video:
Knowledge discovery
Once you’ve created your semantic model for your enterprise information architecture you can dive deeper into your data and unlock insights and discoveries that help drive business decisions. With metaphactory, you can:
-
Identify early on the business processes affected by a system shutdown, such as CRM downtime, to mitigate the impact on processes like invoicing and marketing automation.
-
Map and visualize all physical systems through a visual interface to determine which ones would be impacted by a new data protection law in a specific territory
-
Semantic search and conversational AI interfaces can build on the semantic layer and enable any user to easily interact with the data, while receiving high quality, trusted and verified results
Keep your eye out for a follow-up blog, where we’ll discuss how a customer built an enterprise information architecture based on a semantic model.
Try it for yourself
Now that you’ve learned about enterprise information architecture and how to approach implementing one within your organization, it’s time to get started.
Speak with an expert to discuss your organization and specific use case, and learn how metaphactory can support your EIA journey. You can also inquire about a demo or free 4-week trial of metaphactory.
Contact us
