Illuminating the Mean Streets - Investigating & Quantifying Data Quality Problems

Nigel Turner, Vice President, Information Management Strategy, Harte-Hanks Trillium Software

In my previous series of blogs I focused on how one large global organisation, British Telecommunications (BT), recognised the need to take action to address data quality (DQ) barriers to the development of its business, and started to identify candidate cross-organisational DQ problems that could be addressed within the broader context of an enterprise-wide data improvement initiative.

Now, I plan to highlight how this long list of potential problems was narrowed down to a specific list of initial improvement projects, how business cases for action were made, and how these cases won the backing of BT’s senior executives. This particular blog focuses on how the list of DQ suspects was further narrowed down. Key to this was a detailed investigation of the problems.

It was the time to use forensic techniques. It was time to call in the detectives.

One famous fictional investigator, Raymond Chandler’s Philip Marlowe, would have recognised the challenge. As he walked alone through the mean streets of LA, Marlowe would learn that dirty data quality deeds had been done, that unspeakable crimes involving data had been perpetrated. His job was to establish the facts, and put things right. As one of Chandler’s most famous Marlowe novels, the ‘Big Sleep,’ memorably stated, “The streets were dark with something more than night,” so investigation was needed to throw light on the data quality problems, and to point the way to a brighter future.

Like all good detectives our DQ investigators started with what they knew. This was the list of 50+ candidate data quality themes and problems uncovered by earlier analysis. It was now time to identify the prime suspects. But the investigators could not do this alone. Expert help was needed.

So they called in a group that understood the damage these suspects had caused. This group was known as “The Business.” A workshop with these business stakeholders was set up. Many of the business people invited had been identified in the first phase of investigations; others were suggested by them.

The investigators provided the list of suspects and let The Business decide on their fate. They used a combination of established techniques – Force Field Analysis, Benefits Management, and the Boston Grid – to help the Business decide which cases to focus on first. There was much wailing and gnashing of teeth but eventually a shortlist was drawn up. The Business also agreed to meet again when the detailed analysis was complete.

The investigators did what they were told and conducted a detailed examination of the prime suspects. They used two primary tools. First, they interviewed key people impacted at all levels in the organisation, recorded process and people problems caused by the data, and did a first cut assessment of the financial and other impacts (e.g. regulatory compliance) of these problems. In parallel with this, they put their magnifying glasses on the data itself, first identifying the key sources where the delinquent data was held.

Using data profiling tools they then assessed the scale and scope of the data problems in these key sources. They also put together an initial improvement proposal, embracing people, process, technology and the data itself.

Some general rules emerged from this process. First try to express the impact of DQ problems on the business by their financial impact. This is best done by estimating the costs of poor quality data. What processes fail when data is incorrect or missing? How much does each failure cost? How often do these failures happen? What are the estimated costs of failure in six months, one year’s time?

Second, what other damage is done? How do these failures damage the company’s image and brand? How do they jeopardise the organisation’s ability to meet its legal and / or regulatory obligations? How are customer perceptions affected? How is the morale of the organisation’s people impacted?

The end result of these investigations was a detailed analysis of a small number (3-5) of key DQ related problems. The prime suspects had been rounded up and interrogated. The next step was now to decide what their fate would be. It was time to call in the Business again. See my next blog for what happened next...

Comments

[Kirana]

Thanks for the tip, David, sniatrtg to YFD now. Looking forward to getting an invite to see how we can use this in humanitarian contexts.