Fixing a data quality problem, it's only a problem when its breaking the law or unethical.

Be a hero for the CEO and Shareholders

Personally, I always want to make my boss look good.  Some managers are gonna say this is political suicide.  Those managers are probably the cause of the problem.   

Can you imagine working for someone who's slogan was "the truth doesn't matter"?  I've heard horror stories.  I would rather starve then work for a manager with such low standards.  

I have built my career on doing the right things even if they are harder.  If your one of the best, you want people working for you who do their best before what's best for them.   of my managers have loved me for it.  They know, I am able to fix anything and I get it done within the law and the auditors will love it.  

So let me take you through the tricks behind a data quality problem.  How it effects a record and what to do about it.  

 Identify the issues in ERP master records

• Party

• Offer

• Account

  
  All tables with these master records are in scope and you will not find anyone volunteering the information.  Go straight to the person managing the physical tables.
  Do not read standards, do not trust word of mouth.  Get this person and their management chain to understand your scope.  
  
  The accuracy of the persons response will make your project successful or marginal.
  Using the returned information, begin the profiling of these tables in the database

When I want to profile a data quality problem?

• Step 1 understand the problem from the stakeholders perspective
• Step 2 assess the stakeholders role in the data and the process step the stakeholder work within.  
• Step 3 identify the stakeholder group with the role of creating, when the data can be enriched and how the data is enriched?  
• Step 4 assess the variance between the process owner, the stakeholders who would re-use the information and how the information get's to the stakeholders with the quality failure.  
• Step 5 profile the data starting with the application where the data is "created" a physical table should be found with a 1:1 match of the entry points.  
• If not found, a record retention process may be broken.
• Inputs must be captured and written, then validated during the close (more often usually)
• Each validation results in a record, when financial it must be retained in it's original format.  
• Changes must be enriched never replacing the original records.  
• Many restatements are forced and changing the source information

Meeting the Sarbanes Oxley 404 assurance-be a hero for an executive

• When I measure cost savings from waste, I measure the columns with information from the ERP master records by the number of rows returned in the profiling or daily batch report. 
• When I am performing a security architecture role as the person who qualifies the projects and guides the application threat assessment during the SDLC;
• A threat assessment must be done on any create or get process introduced 
• The business intelligence process should not create, while it may present information by joining two data sets, the use of in memory or a physical table writes to a new table with both parts of the information.   
• Any mis-match between the source application - in the master data records we must assume any values created, updated and archived MUST be done with the MANAGEMENT CAPABILITY user group.  
• These management capability users are a team of people who must associate and document a series of steps in due diligence and retain the records for audit purposes.  
• generally legal, corporate finance and an operations person performing the action with both legal and finance validating the action.