Why Businesses Keep Building Data Silos - Even When Everyone Knows Better

Loraine Lawson

To me, one of IT's great paradoxes is that even as it tackles data integration projects, it continues to create new data silos.


Rick Sherman, a consultant with Athena IT Solutions, recently addressed this issue in his blog, The Data Doghouse. He concludes there are two reasons we still have data silos, even when we're striving for that ever-elusive "single version of the truth":

  1. Business intelligence and datawarehousing projects are typically organized on a tactical, project-by-project basis.
  2. There's no overall information architecture to guide that blueprint.


Okay-that's what causes them, but it still doesn't address the why. Why not change to a more strategic approach?


Sherman thought that question through as well, and lists five factors at work in favoring a tactical over a strategic approach:

  1. Group-based funding for projects.
  2. Jargon confusion. "There are many similarities between CRM (customer relationship management), SCM (supply chain management), budgeting and forecasting, performance management and balanced scorecards. Many people do not see that each has data, data integration and business intelligence that can be shared across projects," Sherman explains.
  3. Technology silos-in other words, they treat different integration technologies (ETL, EAI, data virtualization, etc.) as different applications and projects-when it's all just different approaches to data integration.
  4. IT's organizational disconnect, which keeps "data people" from talking to and working with "application people" -- a political problem also noticed by Judy Ko, Informatica's VP of Product Management and Marketing.
  5. No architecture helping to shape tactical projects.


This piece is Sherman's second on the topic of how people, process and politics impact data integration. He's promised more, and at some point I'm sure he'll come back to one solution he's written about previously: Integration Competency Centers.


Informatica, for whom Sherman sometimes writes, is huge on the concept of ICCs. In fact, the data integration company recently commissioned a study on the the potential ROI of Integration Competency Centers. Forrester, another proponent of ICCs as a best practice, conducted the research.


Forrester concludes ICCs can pay off in a big way. It found running an ICC can cost about $2.7 million over a five-year period, but the typical enterprise can save more than $5.5 million in IT and business benefits over five years. So, for large enterprises, that's double your money back-not a bad ROI.


Alas, there's nothing in the free report about how much an ICC would cost or whether it would payoff for smaller companies.


Joe McKendrick, who wrote about the report's findings for Informatica's blog, noted there's another potential payoff for ICCs: An ICC can help with SOA.

Add Comment      Leave a comment on this blog post
Jun 3, 2009 8:16 AM Rick Sherman Rick Sherman  says:


Thanks for reading my blog and for your thoughts on the subject. I am a techie (or nerd) at heart but people, politics and processes are critical to success. The engineering school where I earned my undergraduate degree had a requirement that engineers had to take a social science or humanities course each semester to broaden our education (fyi: I got a minor in Political Science). In a similar fashion I think IT folks need to examine the 3 Ps if they want to have long-term success in DW & BI.

I will talk about Integration Competency Center (ICCs) next week but will also discuss program mgt; BI CompetencyCenters (BICC) with Gartner as a proponent; and Skills Assessment & Training. I already discussed data governance in this series.


Jun 3, 2009 12:35 PM Rob Eamon Rob Eamon  says:

Another contributing aspect: an on-going focus on data and data integration. Until people stop thinking in terms of "I need the data from XYZ application over here" then the problem will continue.

Service-orientation is ostensibly one approach to get away from that. Instead of thinking about getting access to data, the thinking becomes about what one wants to do. The doing will definitely involve data but in getting to the doing aspect it should become clearer what are the common components.

IMO, an ICC can be useful (if horribly named) but having an architecture team that guides the development and evolution of architecture is more important. I wonder how many successfully ICCs fill the architecture void, thus it is the architecture and processes that an ICC puts into place which provide the value, not necessarily the integration aspects per se.

Jun 11, 2009 1:48 AM bird bird  says:

software programmers just like different gods create different softwares or languages just like god create tree bird river human just human become god of software ahhahahahhahhaahhhhahh

Jun 11, 2009 2:08 AM bird bird  says:

just like Adam n eve evolve to so many different colors human i think this question is never ending debate every software programmers also say their software or languages is best of best n the problems is nobody understand all languages of software i think u need a guy to study all languages of software i think b4 he or she learn all the languages already evolve to newer species until die also never learn all languages of software that's y everybody dare to say their software languages or software programmed is best since no idiot will buy all software to test which is the best or no one software or language suit all different companies just waste time to debate


Post a comment





(Maximum characters: 1200). You have 1200 characters left.



Subscribe to our Newsletters

Sign up now and get the best business technology insights direct to your inbox.