How do you justify data modeling?

I am writing this blog on the plane ride home after teaching the Data Modeling Master Class in San Francisco. On my mind is a challenging question that was asked by one of the participants in the course: “How do I justify data modeling to my management?”

The reason this is a challenging question to answer is that justifying data modeling to management requires some detective work up front. Specifically, we need to answer three questions:

1. How do they define data modeling? The two key phrases in this question are ‘they’ and ‘data modeling’. Who really needs this justification? Is it the outwardly skeptical manager who needs the justification or is it higher-ups? Is it managers on the business side responsible for providing the funding for data modeling, or is it IT managers who are under extreme time and budget pressures to deliver systems. Is it developers promoting agile methodologies such as Scrum and Extreme Programming? We need to determine who really needs the answer to this question and then we can tailor our response. The second phrase is “data modeling”. In Data Modeling Made Simple 2^nd Edition, I defined data modeling to be “the set of techniques and activities that enable us to capture the structure and operations of an organization, as well as the proposed information solution that will enable the organization to achieve its goals.” This includes the analysis necessary to gather the data requirements, finding out and documenting the meanings of all terms, and the element mappings required if data will be sourced from an upstream application. In some shops data modeling can be mis-interpreted to be just the process of building the pretty diagrams with boxes and lines. It is so much more! By intrinsically linking to analysis and the resulting database structures, it becomes an easier sell – after all, why would we skip the requirements phrase and jump right to code?
2. Where’s the beef? Justification requires a meaty response. That is, justifying data modeling does not require the same more generic (yet accurate) responses for the question, “Why do data modeling?” Explaining why we do data modeling leads to responses such as because we can better understand requirements, build a more robust design, provide greater consistency across projects, do the upfront work which leads to less rework later on, and so on. Justifying data modeling however, requires more precision such as a number (e.g. Return on Investment or Return on Total Assets), or a really good story (e.g. “Remember last time we didn’t do data modeling, and we went $100,000 over budget during development and still ended the project with unhappy business users?”)
3. What is the question behind the question? What is really important to the person needing the justification? Often it is data quality. Therefore the question really becomes “How do I justify data quality?” This is an easier question to answer. There are lots of numbers out there proving the value of data quality. There are also some really good stories showing disastrous results when data quality issues occurred. Once while working at a manufacturing company I justified data modeling using this data quality angle by bringing up a story about how poor data quality led to the loss of millions of dollars and loss of invaluable credibility at this manufacturing company’s rival. Just by searching the web, you can find stories such as this.

You can see why justifying data modeling can be challenging. The short answer to “How do you justify data modeling?” is always the typical consultant answer (that I use quite a bit by the way): “It depends.” But behind this two word response are these three questions to answer first! Until the next blog…