The following is an article written by Trellance’s Associate Vice President of Data Consulting, Merrill Albert. The article originally appeared on CUInsight.com
Dashboards are an important tool for uncovering potential data quality issues. Data quality is a key issue for most organizations today that rely on data to understand customer behavior, dissect trends, and make predictions. Implementing data dashboards provides an at-a-glance visualization of a credit union’s potential issues and helps monitor improvement progress.
As helpful as they are, dashboards alone aren’t a data quality remedy. It’s important for credit unions to understand what they can achieve from dashboards, and what the limitations are. To get the most out of yours, keep these four principles in mind.
1. Always Examine the Causes of Data Quality Problems
Your dashboard will provide an important window into data quality issues. It can help you identify problems, but your job is to solve the problem at its source. This is the only way to make lasting data quality improvements.
If employees are entering data incorrectly or incompletely, provide training so they understand the importance and course-correct. If a vendor is sending data incorrectly, establish clear standards so they can do it right. Your goal is to avoid spending your team’s time fixing the same data quality problems again and again. It’s not satisfying work for anyone, and prevents them from spending their time on more high-value efforts.
2. Revisit the Rules
Your data quality dashboard shows how your data aligns with your business rules. For example, if your rules state that member addresses must have a zip code, the dashboard will show how many records are missing zip codes and which ones.
Business rules play an important role in assessing and addressing data quality, so they deserve their own attention. Ask: who created these rules? Are they the default rules provided by the data quality tool vendor, or were they developed by a single person or department? Could the rules be too superficial? The most beneficial rules will be developed by a cross-functional team that includes IT and business managers representing all business functions. The collective perspective will offer more comprehensive views of what the rules need to accomplish.
3. Give Your Metrics Meaning
Many dashboards use a graph or speedometer to pictorially show you how good your data is. It’s nice to get a quick read on data quality, but what does this number really mean?
If your dashboard says that your member data is 90% “good” – that sounds pretty fine, right? But, it’s important to look under the hood of that percentage in order to understand what it really means. Ask the following questions:
- Is 90% an average? Are some individual data quality checks higher or lower?
- Does that 90% contain data quality checks across the disciplines of accuracy, completeness, conformity, consistency, integrity, timeliness, uniqueness?
- What’s causing the problematic 10%?
This context will make communication about your data more meaningful.
4. Use Your Dashboard as a Communication Tool
Dashboards are designed to make information easier to understand and digest, but they won’t do all of the communication for you. Just sharing the dashboard with team members likely won’t be enough. Use them to provide a focal point for clear explanations and constructive conversations.
If your data quality is less than 100%, there should be regular communication about what is introducing errors and what should potentially be used with caution. Staff who use data as a regular part of their work should fully understand data problems in order to understand the potential impact on the results. Bringing the right people into the conversations enables broader, more effective solutions.
Member Addresses as an Example
Let’s take a simple example using addresses. Does your dashboard define addresses as “100% good” because every field is populated? Sometimes, if staff hasn’t been properly trained, they may take shortcuts when entering data. For example, that could mean entering “11111” into a zip code field to move the process along faster because it’s a required field. By this measure, you couldn’t consider the data to be 100% accurate or valid, so it’s important not to derive false comfort from the metric.
Maybe the dashboard says your addresses are 90% good. Before automatically accepting 90% as good enough, you’ll want to know which elements impact the 10% and weigh the pros and cons of investigating and fixing them. For analytics purposes, perhaps zip codes are the most important element, and you’re confident those are accurate. The 90% on your dashboard may be enough.
But, maybe your credit union invests heavily in direct mail. The dashboard helps you understand that 10% of your mail may be returned. What if you have a required communication, such as a data breach notification. Would 90% still be good enough?
In Conclusion
Dashboards are a helpful tool for providing quick insights into the quality of your data. But, your analysis and understanding shouldn’t end there. Dashboards can clue you into what data issues may exist so you can solve them before data problems become business problems.
Merrill Albert is the associate vice president of data consulting at Trellance.
