4 of 7: Deadly Cliché #4 - Data Quality is the Key to AI Success

Introduction

Overused clichés can become stale and important context can be lost, diluting their impact. Watch out for these common pitfalls in analytics, AI and generative AI projects.

7 Deadly Clichés in a Gen AI World

Data Trust vs. Data Quality

This article will explore what data quality is and what it can do, what data quality is not and cannot do, and why understanding the difference “matters.” The goal is to highlight the significance of differentiating between data quality and data trust. The importance and absolute criticality of data quality is frequently mentioned in almost every AI-related article and presentation. However, in my 17 years of leading data governance and master data management initiatives, I’ve never heard an analytics or AI leader say they want “more data quality” or “better data quality.” Instead, they consistently emphasize the need for their teams and models to be able to “trust the data.” Data quality and data trust are not the same.

Data trust and data quality are closely related concepts, but understanding the distinction between them is crucial, especially when working with digital transformations, high-value analytic assets, and high-profile AI models. While data quality is necessary for data trust, having high data quality alone does not guarantee that the data is trusted. Data quality initiatives can become so focused on tangible KPIs and metrics that the intangible aspects—like user confidence—can get lost. This mistaken belief—that data quality alone will create trust—feeds into the dangerous cliché that "data quality is the key to AI success." Let’s dive deeper into the concepts of data quality and data trust, and compare the two.

Data Quality

Definition: Data quality refers to the characteristics (or dimensions) of data that make it accurate, complete, consistent, timely, and reliable for its intended use. High-quality data is free from errors, duplicates, and inconsistencies, and is appropriately formatted and available when needed.

Importance:

1. Reliable Decision-Making: High data quality ensures that decisions made based on data are accurate and well-informed. Poor data quality can lead to incorrect conclusions and undesirable business outcomes.
2. Operational Efficiency: Clean and consistent data allows for smoother business processes, reducing the time and resources spent correcting errors or reconciling discrepancies.
3. Regulatory Compliance: Maintaining high data quality is essential for meeting regulatory requirements and avoiding legal penalties, especially in heavily regulated industries like finance and healthcare.

Key Dimensions of Data Quality:

• Completeness
• Uniqueness
• Timeliness
• Validity
• Accuracy
• Consistency
• Reliability
• Relevance
• Integrity
• Confidentiality

Common Data Quality Metrics:

Data Quality Scorecards (used to provide an aggregate and detailed view of data quality and health metrics). Examples of metrics include:

• Number of Data Incidents
• Time to Detection and Resolution
• Data Transformation Error Rates

Data Trust

Definition: Data trust is the confidence users have in the data being believable, accurate, reliable, and fit for purpose. It goes beyond data quality by including aspects such as data provenance (where the data comes from), transparency (how it was collected and processed), and security (how it is protected).

Importance:

1. User Confidence: Even with high-quality data, if users do not trust it—due to concerns about transparency, questionable sources, or security—they may be reluctant to use it in decision-making. Data trust is crucial for ensuring that stakeholders have confidence in the data they are using.
2. Data Governance and Ethics: Trustworthy data isn’t just about quality; it also involves ethical considerations, such as ensuring data is collected and used in compliance with privacy laws and ethical standards. This builds trust not only internally but also with customers and regulators.
3. Strategic Initiatives: For digital transformation, AI, and advanced analytics initiatives to succeed, data trust is essential. Users must trust that the data feeding these systems is reliable and that the insights generated can be acted upon with confidence.

Comparison: Data Quality vs. Data Trust

1. Scope: Data Quality is a technical attribute focused on the condition and metrics of data. Data Trust is broader, encompassing data quality along with other factors like believability, data security, provenance, and transparency.
2. Impact: Poor data quality directly impacts the effectiveness of operations and decisions. Lack of data trust means that even high-quality data may not be utilized effectively because users are hesitant to rely on it.
3. Building Data Trust: Data Quality is foundational for Data Trust. However, trust also requires strong data governance practices, transparency in data management, dependent processes/workflows and assurance of data security and privacy.
4. Perception vs. Reality: Data Quality is an objective measure of data characteristics. Data Trust is subjective and based on user perceptions, influenced by their experiences, and the organization’s data practices.

Conclusion

While data quality is crucial for ensuring that data is accurate and usable, data trust ultimately determines whether that data will be relied upon and acted upon by users. High data quality is a prerequisite for data trust, but trust goes beyond just quality—it involves confidence, believability, transparency, governance, and security. It’s important to maintain a clear distinction between the two. For organizations to fully leverage their data assets, they must grow into a unified data strategy which includes data governance and data quality to build strong data trust among users.

?