Getting Started with Data Quality: Tips to Overcome Common Hurdles

At Moran Technology Consulting (MTC) we often work with clients on establishing a repeatable data quality process ahead of an ERP implementation. Even though poor-quality data can result in costly delays or lead to bad data being brought into a new platform we regularly see concerns about commencing a data quality program to proactively address issues. This article covers common obstacles we have encountered across clients trying to start a program, and describes approaches we recommend to mitigate them. We hope this will benefit others who are commencing their journey to data quality utopia.

Introduction

Despite data quality being a cornerstone of data governance, institutions continuously struggle with establishing a program to identify, track, and manage their quality items. In the current higher education environment of declining enrollments and rise in non-traditional offerings, institutions are being pushed to leverage their data more and more. Add in the rise of artificial intelligence (AI) capabilities that depend on high quality data and more than ever there is a need for quality across all institutional data assets. Quality should absolutely be a key metric for all executives in the institution. However, many schools establishing data quality initiatives will face familiar roadblocks which can tie up critical resources and pull the focus away from the goal of leveraging data to achieve the institution’s strategic objectives. In this article, we’ll explore common challenges and offer practical steps to work your way through to establishing a successful data quality program.

Common Obstacles

Establishing a data quality program takes coordination, buy-in, and constant reinforcement as to the benefits of the endeavor. Be prepared to identify and tackle these common hurdles that prevent institutions from launching a successful data quality program. Below are some of the most frequent issues along with how this issue might arise in conversations.

A graphic showing six common hurdles that prevent launching a data quality program:

- No data governance program
- Definition abyss
- Data quality ambiguity
- Tool dependency
- Analysis paralysis
- Master data management spaghetti

Data quality initiatives require stakeholders from various groups so we cannot simply dismiss concerns but rather put together a plan to mitigate the concerns so that the program can move forward. Let’s break down each item and provide some recommended actions that can help you overcome these issues.

No formal data governance program

Picture of a table covering:

Problem Statement: Attempts at establishing data governance have failed in the past so the institution has no data governance program to spearhead the data quality program. Data governance often fails due to a lack of executive buy-in which is often driven by unclear benefits and outcomes.

Mitigation Plan: 1. Identify one area that is impacted by data quality, often student records or institutional research can be great first allies.
2. Define the item and establish a measurement of current quality .
a. Example: Count of employees with future birthdates.
3. Launch pilot providing the business unit with impacted records to cleanse.
4. Measure and document progress . This can be as simple as a spreadsheet with item, date identified, initial measurement and current measurement. (Don’t overthink it!)
5. Leverage results and area's executive support to get widespread buy-in.

Definition Blackhole

A picture of a table with the following:

Problem Statement: The data governance program has launched, and the team is still documenting institutional definitions at a pace of a few times per week. Rather than moving onto attaching the definitions to quality standards, there is a focus on getting everything defined first.

Mitigation Plan:
1. Choose an already defined term or set of terms.
2. Define how data quality could be measured for the term . Be specific, as ambiguity leads to misunderstanding.
a. Issue: Duplicate Suppliers
b. Measurement: Count of sets of duplicate active suppliers based on tax ID
c. Measurement Method: Supplier System Report (SUPP-0015)
3. Measure the quality of the data and discuss results.
4. Establish corrective actions . (Clean it in the system, data patch, process correction…etc.)
5. Continue to iterate through the defined terms, while periodically reviewing measurements to assess corrective actions.

Quality Definition

A picture of the table with the following:
Problem Statement: Quality means different things to different people and is ambiguous without proper context. Institutions can get caught up in debating what constitutes a data quality issue.
Mitigation Plan:
1. Leveraging industry standards such as those in DAMA’s “Data Management Body of Knowledge”. Hold a first session to discuss the dimensions of data quality and avoid trying to dig into any specific issue. This will act as a common foundation to build upon.
2. Go through the dimensions again, this time discussing examples of each at the institution, documenting each to begin forming a data quality backlog. (issue name, description, area, contact, comments…etc.)
3. Spend subsequent meetings qualifying data quality issues identified with logic, measurement process, and potential corrective actions.
4. Continue through the backlog and determine a cadence to reassess each item to gauge effectiveness of the actions.

Tool Dependency

A picture of the table with the following:

Problem Statement: Institutions can get enamored with new tools on the market that can “automatically” assess data quality, causing them to think that they should not start a data quality program prior to purchasing a platform.
Mitigation Plan:
1. Open an excel document and create a backlog template for capturing data quality items.
2. Once an item is identified, determine how an existing tool or process could measure the quality . (Application Report, SQL Query, Data Warehouse Report…etc.)
a. Automate these wherever possible to reduce manual efforts and ensure consistency in measurement.
3. Perform regular reviews of the process to document what has worked well and what has not.
4. When the team has iterated through the process on multiple items then assess whether a tool is truly needed and what features would/should be prioritized.

Analysis Paralysis

A picture of a table with the following:
Problem Statement: Data governance meetings are going on and discussions around data quality are commencing but pen never gets put to paper as people want to have a perfect process documented and all unknowns known before getting something going.
Mitigation Plan:
1. Say out loud, "perfection is the enemy of progress."
2. Focus on getting just a single item documented and measured.
3. Any decision on tool or process should be decided within the session with minimal follow-up actions.
4. Start a future improvements document with suggestions from the team as the process moves forward.
5. Agree to revisit the process every 5th meeting to discuss what could be improved .

Master Data Management Spaghetti

A picture of a table with the following:
Problem Statement: Teams understand that there are key systems that manage the same data and get out of sync. This is often viewed as purely an integration issue which becomes a perceived pre-requisite to getting work started. This is what is what is referred to as Master Data Management (MDM) Spaghetti.
Mitigation Plan:
1. Leverage any documentation about the flow of data. (Architecture Documents, Integration Inventories, Integration Specifications…etc.)
2. Validate and enhance the documentation to show the flow of datasets between the systems, prioritize based on importance.
3. Measure the quality of each dataset by comparing against corresponding records in the other system (student address vs student address; employee demographic vs employee demographic...etc.)
4. Utilize log of identified issues to discuss corrective actions .

Conclusion

Achieving a utopian state of data quality – a centralized log, automated measurement, and alerting – is undeniably challenging, but it is achievable if done in a collaborative, iterative, and focused manner. Data quality is a problem that has no silver bullet, as all institutions are different, and will face their own unique challenges. Rather than scouring the market for a product that does not exist, focus your energy on seeking the initiatives that benefit the most from data quality, such as an HCM, Finance, or Student Information System implementations. These kinds of well-funded initiatives will see direct risk mitigation if data quality can be identified, tracked, and managed prior to and during the project. Data quality efforts can save millions in unnecessary delays and prevent the kinds of catastrophic challenges that arise from loading poor quality data into a brand-new platform.

Data governance in general is not an easy task, as it is full of ambiguity and can lead to more circular conversations than most topics. With that in mind, start your data quality program with small manageable steps. Celebrate and publicize even the smallest successes. As you begin to gain interest and broaden the scope you will not only be improving the quality of data, but also uplifting data literacy across the institution. It will not be easy, but the entire institution will benefit from establishing a data quality program.

———

Moran Technology Consulting is an experienced and proven consulting services provider to higher education. MTC offers a full range of IT and management consulting services. Our consultants have worked with over 320 institutions and conducted over 850 projects across 40+ states and 12 countries.