← Back to articles

Business Case: Data Quality - The Missing Process

Grzegorz Gruszka3 min read

Introduction

Let me start with a question: what is the weakest link in a corporate data quality process?

Many answers will focus on the technical side, for example: “data is hard to assess”. That is certainly true, especially when data volumes call for Big Data techniques and infrastructure. There is no doubt that this is a large project, one that requires highly skilled specialists and modern infrastructure.

So is that the right answer? No. Give me two minutes for a brief introduction to the data quality process, and the real answer will be easier to accept.

Discovery in the Data Quality Process

The process has several steps, and most of them are fairly obvious once you start thinking about data quality. Let me focus on the two most important ones. The first is awareness: an honest assessment of what poor data quality actually costs the organization. In my experience, people are aware of these consequences, but at very different levels. Data stewards mostly see data quality as the issues they run into while processing quotes, each of which adds precious time to handling a single case. Back-office process leaders hear about data quality when incorrectly entered data has a major impact on an insurance policy and the client or the front office spots it (before or after it turns into a financial impact, such as a large payout or a coverage gap affecting other institutions). Speaking of other institutions: regulators and government bodies require insurers to manage data quality properly. Regulation is a last resort, but sometimes the government has to step in to fix problems in an industry. That is why, sooner or later, top management gets involved in the company’s data quality process too.

Refinement and Continuous Improvement

So where is the problem? How do you deal with so many different expectations and no big picture? How do you prepare a Data Quality Process Owner for this role, when the process has to cover everything from the bottom to the top - from the data steward typing data into forms (or the client filling them in directly, with a different understanding of the data definitions than the company’s) all the way to the ETL processes that load the data warehouse behind your analytical applications? There is no simple answer.

The second step is data remediation. I have been in many discussions where someone proposed a single, simple fix for every data quality issue: “Let’s validate data at the point of entry” (in client forms or in the operational tools the company currently uses). What follows is always a similar list of objections: we have more than one system; these checks will kill system performance; we don’t own the source code needed to make the changes; the technical implementation will cost a fortune (in fact, poor data quality already costs the organization money - it’s just that nobody in those meetings knows how much); and so on. I’ll stop there.

Even with data monitoring in place and a perfectly prepared exception report (showing the incorrect and expected values, with a clear pointer to where each one should be corrected and by whom), fixing data in the source system is far from easy.

Now I can finally reveal the answer. Organizations don’t know who the actual data owner is. Processes are highly distributed, forms were filled in by employees who have since left, or we can’t even tell what the correct value is, because nobody is going to go back to a client about a policy that has already been closed.

If you are starting to think about data quality in your organization, let us help you with our experience. My team will take care of this process end to end. We will join the discussion with your Board of Directors, your top management, or the technical teams who run your business applications.

PS. There is no better dashboard for monitoring data quality KPIs than a DQ heat map. It is a single tool for presenting data quality insights and a quick health check of the company’s critical data elements. Here is an example:

Read next

400 PLN

Free consultation

A consultation worth 400 PLN for your company - free of charge. You pick the date.

Pick a date (opens in a new tab)

or leave your details - we will call you back

Fill in the form to download the resource

Enter your full name and e-mail and the download link will appear right away.