/ The Case for Moving to a Domain-Centric Data Architecture

This blog, originally published in September 2021, was updated in July 2023. 

Domain-centric data architecture is very closely related to data mesh and shares two of data mesh’s pillars: domain-oriented design and federated data governance. For organizations that are interested in the benefits of data mesh but don’t want to fully commit yet, a domain-centric data architecture is a solid step towards data mesh. 

As someone whose role at Domo is to provide data governance advice to the company’s largest customers, I have lots of conversations with IT leaders about data lakes—and how Domo’s data experience platform is capable of querying data directly from those lakes. 

At its core, a data lake is a centralized repository that stores all of an organization’s data. That data can be structured and unstructured but is typically separated by domain and controlled by a team often called the Center of Excellence (CoE). 

Organizations I speak with tend to already have a data lake—whether it’s in the cloud or on-premise—or are looking to implement one in Domo. 

The reasons for wanting to implement a data lake are simple: they cost less than data warehouses or databases, are a lot more accessible to the entire organization than data that lives in silos, and are ideal for machine learning. 

What’s more, data lakes make it easy to govern and secure data as well as maintain data standards (because that data sits in just one location). 

For these reasons and others, data lakes are popular investments. But as most data practitioners know, some data lakes have downsides.  

For one, the effort and cost to maintain a data lake and its respective data pipeline can increaseover time. For another, as a data lake grows in complexity, it can lead to delays that impact innovation. 


Better BI with data lakes 

To get better, more relevant BI insights from their data lakes, many Domo customers are moving towards a domain-centric data architecture, whose key component involves taking data stewardship tasks away from the CoE (or IT) and giving them to the business units (BUs) themselves. 

This allows the CoE to focus on the administration, maintenance, and innovation of the data platform while allowing BUs—whether it be the HR department, the finance division, or any other part of the organization—to captain their own ships, so to speak. 

How do you open this door? There are three keys: 
1—Empower BUs to own and look after their own data  
The easiest way to empower a BU is by starting with an individual within the unit that has a good understanding of their group’s data as well as the subject area, knows how to properly use the data platform, and is comfortable answering colleagues’ questions about data or the platform. 

Once you’ve identified who that person is (or could be), getting the entire team up to speed becomes so much easier. In my experience, the best method for teaching users a new platform is through data implementation. Pick a valid-but-not-too-complex use case, then run the use case implementation from end to end. 
2—Ensure that adequate communication channels exist between the CoE and BUs  
This is especially important when you get to the point where various entities are managing content creation and its use. 

One way to create strong communication channels is to form a governance council that a) includes the data leads from each of the BUs as well as a representative of the CoE, and b) is willing and able to meet regularly to discuss platform usage, lessons learned, and opportunities for data sharing. 

Another way is to build a custom landing page on the data platform that has recent news, any rules or regulations around data use, and links to training so that your BUs can keep their skill levels high.  
3—Create a platform-monitoring dashboard  
When you get to the stage where you’ve got a lot of new content creators on your data platform, you will need to keep track of them to ensure they are playing by the rules. That’s where a platform-monitoring dashboard can really help. 

In Domo, the primary source of this data will be the Activity Log dataset. This dataset has a list of all actions taken by any user. Some great visualizations you can start with in your monitoring dashboard are: 

  • Policy adherence rates. When implementing Domo, decide on a naming convention and stick to it. A visualization showing naming convention non-compliance will help ensure users adhere to this policy. 
  • Data exporting. Users who export data from your data platform represent a security risk—and could mean that users don’t understand the platform well enough and are exporting the data for use in other data tools such as Excel. A visualization of this can be used to identify these security risks and areas where more training is needed. 
  • Content creation benchmarking. A visualization comparing each BU’s content creation rates could be used to identify where training or adoption is insufficient in a BU. This information could lead to more training workshops to increase adoption. 

What a domain-centric approach can do for you 

With a domain-centric approach to data architecture, a whole new world within the modern BI universe opens, because the following takes place:   

  • Effort is distributed to the correct parts of your organization. As your organization’s BI implementation continues to grow and spread across the enterprise, the effort to maintain existing content—as well as create new content—will increase. Enabling and empowering BUs to build their own content can drastically reduce the need for the CoE to be involved in that concern, which in turn enables the CoE to focus on making sure that increased usage doesn’t become a problem area. 
  • Content ownership is assigned to those who know it best. A domain-centric data architecture allows the finance team to control finance data, the HR team to control HR data, and so on. Not only does this reduce the security risk of the CoE accessing potentially sensitive information, but it also ensures that the users who are most familiar with the subject are the ones managing, creating, and validating data. 
  • Shadow IT becomes less of a possibility. Something I’ve seen with customers who restrict content creation to only a few users is the development of shadow IT, which is a term for when systems are deployed by departments to work around the shortcomings of systems they’ve been asked to use. Shadow IT operations can appear benign. However, as they grow, not only do governance risks such as security and regulatory non-compliance, lack of auditability, and data leaks also grow, but the possibility of a political battle between the new shadow IT and the existing data platform becomes more likely, which can negatively affect company culture. 

To learn more about implementing a domain-centric data architecture as part of your modern BI strategy, contact me at karl.altern@domo.com. Or, if you are already a Domo customer, contact your account executive (AE) or customer success manager (CSM). 

NOTE: This post was written with guidance from TreeHive Strategy principal Donald Farmer, an internationally respected speaker, writer, and consultant who served as moderator of Domo’s ‘Curiosity: Do Data Differently’ video series in 2020. 

Try Domo now.

Watch a demo.