Three critical data quality mistakes and how to avoid them

A minor miscalculation in their accounting system, connected with terms of service update, resulted in a data glitch. It wasn’t a hacker attack or any other big event. Uber’s system just deducted taxes before, not after charging their 25% commission. The cost? $900 per driver in repayments. Times hundreds of thousands of drivers. Plus endless hours tax errors reconciliation. And the craziest part? Nobody noticed until they manually updated those terms again, years later.

If data quality mistakes impacted Uber, they may impact also your company. Why? A study by Experian found that 94% of organizations suspect their customer data inaccurate.

Let's explore 3 critical data quality mistakes that businesses make. And how you can avoid them.

‍

Mistake #1: siloed metadata management across tech stack

Fragmented metadata across different systems makes it impossible to maintain consistent data quality.

The tech challenge: when business glossaries, data catalogs and rules exist in separate systems, companies can't implement a complete data quality management.

The real impact: Accenture’s report shows that unified metadata management can reduce time to insights by 50%. This translates to an average of $8.2 million in annual productivity gains for large enterprises.

How to avoid this mistake:

• Implement a unified metadata management platform and standardize exchange formats.
• Create automated metadata harvesting from all systems.
• Establish clear metadata ownership and governance.

Goldman Sachs investment in unified metadata management led to a 42% reduction in data quality incidents. By connecting their data catalogs with quality metrics, lineage information, and domain knowledge, they created a self-improving system that automatically identified and remediated issues across their data landscape.

‍

Mistake #2: insufficient data lineage tracking

Modern data ecosystems are really complex. Data travels through multiple systems, goes through numerous transformations, and gets combined with other datasets. Without proper data lineage tracking, it becomes impossible to identify where quality decreases.

The tech challenge: most companies track data lineage at a high level only (system-to-system). But miss the column-level and transformation-level lineage that's essential for full quality management.

The real impact: McKinsey’s study found that banks with top data lineage technology were able to solve data quality issues 65% faster than those without. Saving an average of $4.5 million per year in reduced analysis time.

How to avoid it:

• Implement data lineage technology that captures metadata at each transformation stage.
• Create lineage visualization dashboards that business users can actually understand.
• Set data quality checkpoints at key transformation step, map complex data relationships.

’’Data-lineage tools have traditionally obliged companies to use a tool from their data warehouse vendor. Today newer technology can operate across multiple data platforms’’ says Tony Ho, the author of McKinsey’s report.

‍

Mistake #3: ignoring data quality dimensions beyond accuracy

Most companies focus only on data accuracy. And ignore other critical dimensions needed to manage data quality

The tech challenge: true data quality includes at least six dimensions: accuracy, completeness, consistency, timeliness, validity and uniqueness. Each dimension requires dedicated measurement approach and strategy. This needs to be tracked in one place, end-to-end.

The real impact: TDWI in their 2024 State of Data Quality Report explains that only one in ten companies have complete and automated data quality management. Other 90% carry data quality risks.

How to avoid it:

• Define metrics for each data quality dimension relevant to your business.
• Implement one technology to monitor them all.
• Create quality scorecards. And use them also for new processes.

Data quality management is an ongoing journey. It is difficult to follow without a dedicated team and clear metrics to manage. TDWI underlines that only 14% of enterprises have implemented operational tools that enable end-to-end data quality monitoring.

‍

To summarize

These data quality challenges require complex tech, but the underlying principle remains simple: you can't build reliable insights on unreliable data. As Tom Davenport, MIT professor, writes in Harvard Business Review: "In a world obsessed with the next big thing, we've somehow forgotten the first big thing: good data.’’ Because at the end of the day, your data is your business. And if your data is broken, your business is broken too.

Want to take your data quality to the next level? Explore Black Tiger Framework.

Written by Michal Kolataj

‍