A Data Science Central Community
For many non-technical individuals in the business world, data modeling can seem like a strange and somewhat terrifying realm. Even those who are data-savvy and regularly consult and analyze data in their day-to-day operations will often view modeling as perplexing under-the-hood stuff that is best left to data analysts or IT staff.
To an extent, there’s some truth to this: advanced data modeling can quickly become a complicated affair (although the right business intelligence software makes it much, much simpler) and is often best left to the pros. However, even if you’re a non-techie who isn’t going to be the one actually creating the company-wide data models, having a rudimentary understanding of the basic concepts can help you, the data analyst and the business achieve the best results from the BI processes in place - and here’s why:
A statistic that’s cited to the point of cliché states that analysts might spend up to 80% of their time preparing data for analysis. Within the data preparation process, in addition to cleaning and normalizing data, creating the data model or models typically takes up a large portion of that effort. More importantly for our purposes, it will also define the types of analyses that can be performed, and consequently the types of dashboards or reports the end users will be able to view. This slideshow summarizes the importance of data modeling in business intelligence:
This means that to truly understand the capabilities and limits of your BI system, you need to have at least a rudimentary concept of the way your data is structured, from a conceptual viewpoint. This doesn’t mean being able to decipher every SQL query running in the background - but it does help to have knowledge of how your company’s data is gathered, stored and managed, and how it all connects to your business goals.
As mentioned, you might not actually be the one working with data models in your organization. However, as long as you’re analyzing data - even completely passively, as a recipient of weekly reports - the data model in place affects the type of data you’re seeing and the conclusions you can draw from it. Being able to effectively communicate business rules to data modelers, and to understand from them what they need in order to make the data comply with these business rules, can do wonders to improve the quality and relevance of data being used in the organization.
Business intelligence is ultimately meant to serve the business and to make it easier for executives to make data-driven decisions. The ability to clearly communicate requirements to BI project leaders is a crucial means to this end.
Assuming you want to do more than just consume static reports - i.e., leverage the capabilities of modern analytical tools to explore data and perform ad-hoc analysis - then you'll have to realize how the data model in place affects the types of questions you can ask when querying your data. By grasping the fundamental logic which connects your various data sources and tables you'll be able to enhance your ability to analyze data independently and produce more meaningful insights.
While the more ‘hardcore’ data modeling is indeed still best left to professionals, it’s important to realize that today’s self-service BI tools give you extensive DIY capabilities - even when you’re working with data from multiple sources. For a somewhat self-serving example, Sisense provides simplified data preparation, enabling you to naturally connect data sources according to easily identifiable common keys - without any scripting or coding, or the complications associated with systems that rely on OLAP or star schema for database design.
This means that you can do a whole lot of data modeling and analysis completely independently - without ever bothering the professional data analysts or IT department in your company. This in itself is a great reason to gain some basic data modeling chops and take your first step towards data heroism.
Business executives might become frustrated when analysts inform them that an ostensibly simple analysis can’t be performed within the current systems, or that it might take longer than expected. Many of these issues arise due to data modeling problems, but more than often they are solvable - for example, modeling might be needlessly complicated due to the need to adjust the model so it can work with a hodge podge of different tools, many of which can be replaced with a single-stack BI tool. Other times there are fixable issues around the way data is being collected or stored. Even if there’s no quick fix in sight, it’s important to know where you stand to help you allocate your analytical resources in the way that best promotes your business’s interests.
Data modeling essentially defines the relationships between various tables and databases. If your organization is still working with spreadsheets, this may seem less relevant - but odds are, it isn’t. As data grows more complex, an increasing share of companies find themselves regularly relying on cross-database analytics, with data originating from many disparate sources.
In this state of affairs defining the relationship between these different data sources and the structure of your analytical repository becomes more important than ever - since these factors will play a much larger role in the way your company handles BI, and it’s likely to happen sooner rather than later.
Finally, and in addition to all of the above - it never hurts to become more data literate. After all, it’s a data-driven world, and every day data is becoming a more important factor in the regular operations of nearly every business imaginable. With all other things being equal, the person who has data on his or her side is always on the right side of the argument; and being more data-savvy will almost certainly make you better at your job, particularly on the managerial level. In other words - enriching your knowledge of all things data and analytics is great for your career (not to mention your general knowledge).
This post was originally published here and has been republished with permission.