Tuesday, August 30, 2011

Top Ten Root Causes of Data Quality Problems: Part Four

Part 4 of 5: Data Flow
In this continuing series, we're looking at root causes of data quality problems and the business processes you can put in place to solve them.  In part four, we examine some of the areas involving the pervasive nature of data and how it flows to and fro within an organization.

Root Cause Number Seven: Transaction Transition

More and more data is exchanged between systems through real-time (or near real-time) interfaces. As soon as the data enters one database, it triggers procedures necessary to send transactions to other downstream databases. The advantage is immediate propagation of data to all relevant databases.

However, what happens when transactions go awry? A malfunctioning system could cause problems with downstream business applications.  In fact, even a small data model change could cause issues.

Root Cause Attack Plan
  • Schema Checks – Employ schema checks in your job streams to make sure your real-time applications are producing consistent data.  Schema checks will do basic testing to make sure your data is complete and formatted correctly before loading.
  • Real-time Data Monitoring – One level beyond schema checks is to proactively monitor data with profiling and data monitoring tools.  Tools like the Talend Data Quality Portal and others will ensure the data contains the right kind of information.  For example, if your part numbers are always a certain shape and length, and contain a finite set of values, any variation on that attribute can be monitored. When variations occur, the monitoring software can notify you.

Root Cause Number Eight: Metadata Metamorphosis

Metadata repository should be able to be shared by multiple projects, with audit trail maintained on usage and access.  For example, your company might have part numbers and descriptions that are universal to CRM, billing, ERP systems, and so on.  When a part number becomes obsolete in the ERP system, the CRM system should know. Metadata changes and needs to be shared.

In theory, documenting the complete picture of what is going on in the database and how various processes are interrelated would allow you to completely mitigate the problem. Sharing the descriptions and part numbers among all applicable applications needs to happen. To get started, you could then analyze the data quality implications of any changes in code, processes, data structure, or data collection procedures and thus eliminate unexpected data errors. In practice, this is a huge task.

Root Cause Attack Plan
  • Predefined Data Models – Many industries now have basic definitions of what should be in any given set of data.  For example, the automotive industry follows certain ISO 8000 standards.  The energy industry follows Petroleum Industry Data Exchange standards or PIDX.  Look for a data model in your industry to help.
  • Agile Data Management – Data governance is achieved by starting small and building out a process that first fixes the most important problems from a business perspective. You can leverage agile solutions to share metadata and set up optional processes across the enterprise.

This post is an excerpt from a white paper available here. My final post on this subject in the days ahead.

No comments:

Disclaimer: The opinions expressed here are my own and don't necessarily reflect the opinion of my employer. The material written here is copyright (c) 2010 by Steve Sarsfield. To request permission to reuse, please e-mail me.