If you want your data model to be really helpful … Ask these four questions
23 August 2018 Author: Todd Heather
Data modelling sometimes seems like an arcane art practiced by lore masters and mages, schooled in their mystical and mysterious practices in ivory towers with little relevance to here and now pressures of modern organisations.
Nothing could be further from the truth.
In fact, data modelling can be a critical tool to help solve many of the cost and risk issues that plague us. But data modelling needs to be approached with our feet on the ground.
data modelling - It's not that hard
How about this quote on data modelling:
“A ‘look across’ notation such as used in the UML does not effectively represent the semantics of participation constraints imposed on relationships where the degree is higher than binary.”
Are your eyes glazed over yet?
Mine are, and I’m a data management specialist.
But once you strip away all the abstract, technical language, it turns out this quote is a simple statement about how to best show relationships between things: that’s what data modelling is all about.
Principles of Data Modelling
To assure success in modelling, there are some questions that you need to answer before starting and as you keep moving along:
WHY are we data modelling ?
Here’s 4 reasons why you might be data modelling.
1. If you are data modelling because a consultant said you needed a data model, stop now.
2. If you are data modelling because you have some people who are keen to do some data modelling, proceed with caution.
3. If you are data modelling because you are embarking on business / IT project in an important area of your business, carry on.
4. Or if you are aware of serious data quality issues that you intend to resolve with insight gained from the data models. Go for it.
What are the best practices of data modelling?
Considerations of scope:
For data modelling to be practical and useful it needs to focus on something clearly defined, and fairly narrow in scope. It is possible to model an entire organisation with reasonably modest effort but only to a high level.
With data modelling, as with many things, the angel is in the detail. Top-level conceptual data models are used to establish a big picture but are best used to provide an overall scope to govern a series of deeper dives in specific focus areas.
Don’t try to model deeply and broadly. Pick a scope relevant to your project or issues and focus on that. This will help drive out insights and deliver value quickly. Iterate across your high-level model in priority order.
Do we need to get physical?
Conceptual vs Physical Data Modelling
Your conceptual data model will provide you with an overall view of key concepts in your organisation and their relationships. It will start nudging your organisation towards a common vocabulary and shared definitions. Further insights arise at the logical data model level. This is where data issues, inconsistencies, differing definitions and relationships / structures emerge. The process of working through these issues can be a rich source of value from the modelling work.
Developing a data landscape to understand the lineage of physical data sources can be helpful in exposing issues. But your modelling probably only needs to get to a physical data model if you are driving through to implementation of your business / IT project.
Much of the value of the process can be delivered out of the conceptual and logical models.
What does success look like in data modelling?
Or put another way, how would we know we were done? Data management is not a project – it is a function. No one ever asks, “Are we done with Corporate Finance yet?” So, while your data management work will be ongoing, there are a number of opportunities to demonstrate success along the way. It will be important to celebrate successes to maintain momentum and stakeholder engagement.
Successful results from data modelling can include:
- – Agreement across the business about consistency in the use of terms – this drives reduced cost and process efficiency
– Resolution of data issues such as duplication, poor data quality, the proliferation of reference data, and
– Integration of data architecture with business and applications architecture to help create a more coherent overall enterprise.
In conclusion, answering these four questions will help you take a practical, business-value driven approach to data modelling that will help solve real problems that you are facing, especially in reducing cost and managing risk.
 James Dullea et al An analysis of structural validity in entity-relationship modelling, 2002 http://www.ischool.drexel.edu/faculty/song/publications/p_DKE_03_Validity.pdf
ABOUT THE AUTHOR
Todd Heather is Principal Consultant and Practice Lead at Robinson Ryan, Todd Heather, is a specialist in providing leadership for IT organisations that strive to contribute real value. Previously, for 15 years, Todd held the position of practice lead in a management consultancy spanning North America and Australia and, for 10 years, the CTO of a large government agency.