Regular readers will be aware of a recent survey we ran exploring how organisations use data quality within business intelligence initiatives.

The results are in and they certainly leave a great deal of room for improvement.

Data Quality in Business Intelligence: Survey Results

This survey was completed by 203 respondents from the 3rd April 2009 to the 12th April 2009, the survey was promoted across data quality, business intelligence and data warehousing forums on LinkedIn, the Data Quality Pro membership (1200+) and our Twitter followers.

Here are a sample of comments survey respondents posted regarding their data quality challenges:

• Convince and stimulate co-workers to use the data quality processes. They must face the benefits in order to be convinced using it. DQ must be an easy process.
• Cost of solution, achieving understanding about the risks of poor data quality, defining a 'smart' business case (how to measure the increase of revenue due to the DQ solution ?)
• Creating intelligent metrics to support DQ processes. Lack of data dictionaries / data definitions.
• Get the necessary time/budget approved
• Getting funding, communicating to IT the importance. Getting business stakeholders has been quite easy.
• I have answered as though I were still employed by my former employer - a major bank. The challenges have related to no one division being responsible for customer data but all divisions needing quality customer data. Attempts by technology with strong business sponsorship have been partly successful but as the bank takes over more banks, the islands of data re-appear.
• Interest is low. Without being able to turn the data errors to money (how much single data error costs for the company), there is low interest. => Finding the real sponsor for the data quality work is needed.
• It took us 6 months to get an organizational DM structure and have all divisions in support. Now, under the present economical conditions anything that looks like adding administrative tasks to the divisions is banned. Many quality issues can only be resolved with the support of the divisions. Resolution of issues is therefore taking much more time then necessary.
• Organizational problems. Lack of business ownership. Identification of hot spots.
• The biggest challenge we face is identifying the appropriate source system and obtaining that data from our customers.
• Senior Management and organisational commitment and resistance from IT Department. Have worked in two organisations where IT & BI had political battles over ownership of Data Quality and/or Data Governance.

Conclusions

This survey was completed by 203 respondents so the results need to be assessed in this context.

The relatively small sample size apart, the statistics clearly point to a worrying trend where many organisations are implementing business intelligence without any thought to the accompanying data quality process that is clearly a best-practice in this situation.

Intelligence first, Quality later

Of the 31% who have no data quality process in their business intelligence capability, only 9% had plans to implement one within 3 months. This would appear counter-intuitive, why launch a decision-making capability first and then ensure the quality of the decisions afterwards?

Major Issues Reported, Irrespective of Data Quality Management

50% of respondents cited the fact their organisations have major data quality issues in their business intelligence solution. This is worrying as it implies that although 69% of organisations are creating a data quality capability within their business intelligence service, major issues are still getting through.

Where organisations have a data quality capability we found 44% of respondents were still witnessing major issues caused by data quality defects. Clearly, these organisations are struggling to eliminate problems further upstream in the information chains and if they are causing issues in the business intelligence layer you can be sure they are causing issues elsewhere.

ROI and ownership are key challenges

Looking through the comments we received, it is clear there are some common themes. Creating a business case to obtain funds appears several times as does finding the right sponsor and stewards to eliminate political struggles.

Heavy reliance on scripting points to IT ownership?

The results show that more organisations favour in-house scripts and code than vendor products. What are the implications of this? Lack of re-usability, difficulties in training new staff, loss of knowledge when key resources leave, increased complexity and cost, disconnected "islands" of data quality, lack of governance.

Whatever your thoughts on 3rd party tools vs scripts I think no-one would object to the business taking more responsibility for data quality. How this is delivered is a bigger topic and certainly one ripe for further discussion but there are clearly advantages in seeing less reliance on local scripts for cleansing data quality defects downstream as opposed to root-cause data quality improvement upstream.

It is clear from these anecdotes and the survey results that organisations are struggling to successfully embed data quality in their business intelligence services. Over the coming weeks we will be discussing this challenge with business intelligence and data quality experts in order to provide some practical advice in this area.

What are your experiences of business intelligence and data quality? Have you successfully delivered business intelligence with an accompanying data quality management process? What lessons can you share?

Useful Resources

Data Quality in Business Intelligence Survey

Embedding Data Quality in Business Intelligence Reports: Introductory Tutorial