By Leon Katsnelson
By Susan Visser
By Bernie Spang
By the DB2 Guys
By Fred Ho
By Louis T. Cherian
By Shweta Shandilya
By Lawrence Weber
By Serge Rielau
By Dwaine Snow
Do you have a data model? Do you even know exactly what one is?
It’s an important question. Companies these days are operating in tighter economic conditions, which puts them on the lookout for operational efficiencies anywhere and everywhere. At the same time, data is available in abundance, growing exponentially year over year. But without a structure or business context to give this data meaning, most information resources remain largely untapped. Can you afford to ignore a valuable resource in a competitive business environment?
I didn’t think so. This is where data models come in.
A data model is set of logical and physical structures built on consistent language to help business and IT groups map their data landscapes.
A logical model consists of business entities and their relationships, defined by consistent business language and independent of the physical layer. The physical model consists of all objects of a database implementation, including various constraints and specificities of the database vendor. It should provide enough information to build dependencies between the database objects and allow for estimation and management purposes.
A model is classified into subject areas for ease of use and reference. These classifications aid in showing interdependence of certain elements across the subject areas. For example, these subject areas are from an insurance data model:
About 80 percent of the data modeling requirements of organizations in the same industry are identical. The remaining 20 percent of the model must be customized to handle company-specific product offerings, geographies, and business focus.
There are a few industry-specific models being widely used right now—for example, the insurance industry uses the ACORD, GDV, Teradata, and Oracle models. Telco uses the eTOM (SID), Teradata, and Oracle Telco models. In most cases, vendor models are at least 75 percent compliant with industry-standard models like ACORD and SID. Physical constraints prevent absolute compliance to these models, but most align in principle.
Global data is growing rapidly, and IT budgets are not keeping pace. This harsh reality means that information leaders must get smarter about how they manage data. Imagine a CIO embarking on aggressive cost reduction exercise while seeking an improvement in data quality, reliability, and processing efficiencies. The CEO is also keen to launch new products quickly into both existing and emerging markets. However, making your data estate—including the massive amounts of data generated through social media channels—ready to respond and integrate seamlessly to support these efforts is no mean feat. It’s also crucial to grow your data landscape in a structured manner that supports streamlined integration. The alternative is forsaking business agility for data spaghetti. Organizations must also enable transparent management reporting to meet mandatory regulatory compliance requirements.
Failure to do all of this could mean missed revenue opportunities—or worse, customer dissatisfaction. Leveraging a corporate business warehouse requires a data model at an absolute minimum. Without a model, analytics and business intelligence efforts will go nowhere.
The more numerous your customer segments, the more products you offer, and the more geographies you cover, the more useful a data model will be to your organization. Your data will be highly complex, created by multiple systems in different formats but still essentially carrying the same business information. Similar policy data existing in different formats across your geographies needs to be modeled so policy data can be shared, transformed, and put to use consistently across the enterprise.
A data model can help generate confidence that your data is trusted, reliable, and secure. Why would you not use one?
Creating your own data model is a journey. A model can be created by pulling together existing data assets and expertise, but this approach may lack business best practices. Prebuilt industry-specific data models can offer a quick start, though they require you to map all your existing data landscape into the adopted model. Even these models will have to pruned and preened to make them fit for purpose. Most models are comprehensive beasts designed to suit businesses of all shapes and sizes.
You add value to whatever model you choose by incorporating your business considerations and nuances, which are unique to your organization. Lay out the path to your target data model and rein in expensive application procurements that don’t conform to the enterprise standard. A model is essential to any standard you can lay down.
Whatever path you choose, you should bring together the best data modellers, enterprise data architects, and business data experts to make it a reality. Many models have been created in silos that cannot fit together, defeating the very purpose they have been created for. Avoid falling into this trap by engaging experts in this field.
Spending time to create your own enterprise model will reap rich rewards for all of your data-related initiatives, saving you from making expensive mistakes you cannot afford in this day and age.
Can you live without a data model? Maybe. But would I recommend it? Definitely not.
Let’s get modeling.
DB2 TechTalk: Deep Dive on BLU Acceleration in DB2 10.5, Super Analytics Super Easy
Thursday, May 30: 12:30 – 2:00 PM ET
Big Data Seminar 2013, Featuring Krish Krishnan
June 14 in New York City
marcus evans Pharma Data Analytics Conference
July 10-11 in Philadelphia
IBM Smarter Content Summit 2013
Big Data at the Speed of Business
Broadcast event replay now available
Information on Demand 2013: Early Bird Registration Now Open
November 3-7 in Las Vegas