Business analysts need to be involved in the process to make it less stressful for everyone, and IT and data science need to make the tech more clear.
A good data model takes into account the context of the business process it must address, and then addresses that context with the types of data needed to inform decisions that support that process. But it doesn’t always come that easily.
SEE: Navigating data privacy (free PDF) (TechRepublic)
There has been significant confusion about data models, which for too long have seemed to be the abstract province of data engineers and data scientists. Part of the reason for the confusion is that data models are always being discussed in the technical (or physical) structure of the models. By physical, I mean the technical names of data elements and datasets, the technical names for databases and data transformations, and the jargon of programming languages like R and Python that end users and many IT staff have little or no knowledge of.
This has resulted in a raw fear of data models among the ranks of users and IT business analysts that companies can’t afford—a fear that has impeded the development of data models that address the end aims of the business, and that don’t digress into data technicals that few people get anything out of.
To change this mindset, business analysts must get directly involved in defining data models—but they don’t have to do this work by taking data science and programming classes in their spare time.
SEE: Decisions, decisions: How to choose between so many big data tools (TechRepublic)
Here’s how it can be done.
1. Define the business requirements
What is the business problem that needs to be solved by the data model? Is it an automated loan decisioning process? Or a recipe formulator for best ingredients to be used in a cattle feed for a specific herd?
The business analyst is best equipped to work with users and to visualize the business process and data needed. The analyst can also describe those needs in plain English.
What should result is a logical data model, usually in the form of a bubble chart, that shows the different data needed and an accompanying narrative that explains how the data must be processed.
SEE: 8 best practices for optimizing your analytics reports (TechRepublic)
While doing this, the business analyst remains focused on what the business needs. He or she doesn’t need to be concerned about which datasets, systems, or programming modules, must be used to make the business model happen.
2. Work with IT and data science
Once the logical chart of data bubbles is developed, along with a narrative of what needs to happen in processing this data, the business analyst meets with IT or data science.
These are the people who transform the logical model of the data model into a physical model that defines the data stores, system internals, programs that need to be written, etc., in IT terms.
SEE: 7 big data wishes for 2021: IoT standardization, stronger use cases, and more (TechRepublic)
IT engineers and data scientists require this physical data model to do their work, but the demands on the business analyst are less. The business analyst only needs to have a working knowledge of technical terminology and processes so he or she can communicate at a high level with IT and and serve as a liaison back to the end user to assure that the data model and the development of application stays on course with the business use case.
3. Trialing and installing the results of data models
Once data models and applications are built, it’s time for the end user to trial them. During this process, the business analyst plays a critical role, functioning as a liaison between users and IT/data science. During the process, analytics applications are fine-tuned, signed off on, and then installed in production.
SEE: 7 on-the-ground big data strategies for 2021 (TechRepublic)
Working together isn’t a huge leap
In many respects, the role that business analysts play in data modeling doesn’t substantially differ from what analysts have historically done. Analysts define user requirements for applications, articulate a basic business design, shepherd the process through IT, and ultimately trial and install the app in production.
SEE: Why is machine learning so hard to explain? Making it clear can help with stakeholder buy-in (TechRepublic)
While there might be a little terminology and technology to master for data model discussions with technical personnel, getting to know the fundamentals and the vocabulary of data modeling isn’t daunting.
This should provide encouragement to CIOs and business analysts who must now attune themselves to data modeling and how to best deliver results to the business and instill confidence in their users.