Mastering Data Management

More in Business Process Management

• New ebook: Understanding Big Data
• 37 Definitions of “Employee ID”
• Defining Terms: Agree, or Doom your Project
• Establishing a Valid Business “Need to Know” for Sensitive Data

More in MDM

• Best Practices in Test Data Management: Discovering and Subsetting
• TDWI Chicago Preview: Data Warehousing, Analytics & More
• Who’s Afraid of the Big Bad Data?
• MDM for Chief Privacy Officers: A healthcare provider example

More from Marty Moseley

• IAIDQ Blog Carnival: Data Fitness & Garbage
• Mind the Gap and Do Data Governance Better
• Measuring Data Governance: Metrics that Matter
• Measuring Data Governance: How Do You Know You’ve Made a Difference?

Share this Post

Recent Comments

• ICMCC News Page » Where East Meets West: ehealth in China and the US on Where East Meets West: ehealth in China and the US
• Kim’s Training Podcast on Training as a Safety Net: Kim May
• TDWI Chicago Preview: Data Warehousing, Analytics & More | Mastering Data Management on New ebook: Understanding Big Data

Make sure you're measuring what matters to the business as a whole

Last week's blog post included thoughts on how we can identify fairly discrete sets of data that matter most and those business leaders who will be stakeholders in ensuring the proper care and feeding of those data.

I also covered a bit on how we use data governance to publish declarations of Principles, Policies, Procedures and Business Rules that will define the levels of quality required for those "most critical" data.

So now, I want to address the fourth issue from last week's blog: You must be able to monitor and measure the progress you're making in improving data quality. But while you need to establish goals for data quality improvement and track your progress, you must also ensure that they reflect the organization's corporate goals.

Scenario:

Let's imagine you're a commercial enterprise and have published a principle or two on the value of customer data to achieving your corporate objectives. Remember that principles address why the data are important to your enterprise.

Let's say that you've also published policies on the levels of quality you want to achieve to meet those principles. Policies state what levels of quality, completeness, consistency, uniqueness etc. are required to meet corporate goals.

Let's further assume that inside those policies you've defined business processes/procedures that clearly define the steps to take to prevent and/or remedy data quality issues.

These processes may define the steps to find existing customers before adding/updating existing records - how to look for and prevent duplicates. Or they may define what to do if an anomaly is found from a workflow perspective.

You'll also end up defining business rules which state how you examine a data value and determine if it is correct or not (in context of a business transaction or other accompanying data), and how to alter the value or raise an alert when anomalies are detected.

The question is: How do you measure whether you are meeting your principles' and policies' objectives?

Data stewards will want to measure pretty discrete technical  characteristics in the data, such as percentage of duplicate records, percentage of records with anomalies in certain fields, patterns in dependencies between fields (every time address line 3 has a value that is the last four digits of the account number it's actually the parent account number, etc.).

But a key goal for overall data governance success is to measure how well the overall principles and policies are being achieved. To do that, you need to identify those metrics that the business cares about and tie those into measurements that data stewards will be taking.

When measuring value anomalies, you might want to look for the number of customer records that have been assigned to a "catch-all" business category instead of a specific product or marketing category. Telesales reps are notorious for using whatever fields are present to capture information that is useful to them, unaware that they're wreaking havoc in downstream systems.

An interesting business metric might be the number of new customers who are being assigned to a phantom account owned by a sales rep, to make it easier for the rep to track their sales transactions.

I remember a case where a sales rep apparently was the corporate parent of over a hundred completely separate consumer customers, simply because of how they'd captured their name and phone number on their records for followup and revenue tracking.

Just yesterday, I was chatting with a colleague and she was telling me about a client (a very large network equipment manufacturer) who was utterly shocked to learn that a majority of their business apparently was in the K-12 (education) market, based on the data collected.

Of course, a majority of their business was not in this sector, but according to the data, those transactions were codified that way. But how do these anomalies affect the decisions made at corporate levels on the allocation of resources? Or how revenue is allocated to different business units? How do the impact margin targets by business unit? Or how revenue and growth might be reported to Wall Street, the SEC and other interested agencies?

Business leaders care about business growth, revenue allocation, customer value, profitability and trends on these over time and across geographies and markets. To really drive home the importance of data quality, you have to tie metrics like duplicate records and anomalies in addresses to these corporate measurements.

You can't just measure what matters to the data stewards, since they're usually looking for finer-grained, lower-level patterns. Long-term data governance success comes from tying your goals back to the things that matter to your business leaders and make your organization tick.