People Data Excellence: Driving Quality through Empowerment, Standardization, and Automation

GIGO ("Garbage in, garbage out") is a significant risk for professionals working with data and AI. Ensuring high-quality (people) data is crucial for building leaders' trust in data-driven talent decisions and reducing the need for manual reconciliation. Moreover, maintaining top-tier data quality is essential for the successful implementation of AI and GenAI technologies.

When building Workforce360, IBM's internal people data platform, our goal was to go beyond abstract data quality metrics. We concentrated on initiatives that have measurable impacts on processes and outcomes, enhancing data quality through empowerment, standardization, and automation.

A. Empower: Putting Data & Knowledge in the Hands of Users

As we began developing our solution, it quickly became apparent that many data quality issues were not caused by technical flaws, but rather by user misunderstandings. Data errors were often introduced at the source (e.g., HRMS), and insights provided by our team were frequently misinterpreted by end users, particularly when the data passed through multiple hands. To address this, we prioritized giving users greater control over their data and direct access to relevant knowledge.

• Self-Management: We empowered managers and employees to manage their own data easily and directly through AskHR, our HR chatbot, fostering a sense of ownership and responsibility.
• Data Democratization: We streamlined the process of granting access to people data for decision-makers, reducing the need for HR mediation and ensuring actionable insights are readily available. Today, 84% of data access requests are pre-approved based on the user’s job role and identity.
• Data Literacy: We centralized our data and metrics catalog and directed all inquiries regarding people data to a dedicated internal community. This community, comprising over 2,000 professionals, facilitates knowledge-sharing and expertise exchange. As a result, we’ve seen a 33% reduction in user requests related to data errors, allowing both our people analytics team and our users to focus more on taking meaningful actions.

B. Standardize: Establishing a Unified Approach for data and processes

A second crucial aspect of data quality is addressing the issue of 'multiple versions of the truth,' which often occurs when different parts of an organization, whether within HR or across the company, collect and process their own data for a common purpose. I vividly recall seeing a slide a few years ago that displayed ten different values for what should have been a straightforward headcount metric—astonishingly, eight of these values pointed to the same IT system as the source of truth!

I vividly recall seeing a slide a few years ago that displayed ten different values for what should have been a straightforward headcount metric—astonishingly, eight of these values pointed to the same IT system as the source of truth!

To address this issue, we implemented Common Governance and radically simplified the processes and experiences that deliver these insights.

• Governance: Enterprise-wide governance ensured that people data and metrics were consistent, which in turn built confidence among decision-makers. For more details about our people data governance, please refer to one of my previous article.
• Simplification: By replacing bespoke reports with standardized enterprise experiences, we reduced complexity and ensured everyone was using the same accurate data. This shift allowed us to eliminate 60% of our reports, minimizing data disputes and keeping the focus on driving business results.
• Consistency: We established a controlled approach to deploying new metrics, reports, and insights, maintaining data reliability. A consistent and trustworthy process for updating metrics definitions enabled leaders to make intentional decisions based on a clear understanding of the overall impact of the change, rather than being confined to the silos of a specific domain or report

C. Automate: Enhancing Efficiency Through Technology

Finally, we focused on improving data quality from a technical perspective by automating as much of the validation process as possible. This proactive approach allows us to identify data errors and inform users before they encounter issues themselves, thereby enhancing trust.

• Data Consolidation: We merged people data from over 35 systems into Workforce360, IBM's internal people data platform, eliminating manual errors, particularly when combining cross-domain data. This consolidation was crucial not only for accurate reporting but also for accelerating the development of AI solutions. For example, with Workforce360 APIs, IBM's data scientists and software developers can now incorporate people data into internal applications seven times faster, transforming a once complex process into a largely self-service workflow.
• Data Validation: We implemented automated checks for data and business integrity to ensure that only accurate data enters Workforce360. These checks also detect and address any “metric anomalies” before the data reaches users. Currently, we have several dozen rules that cover various HR workflows, such as organizational hierarchy, attrition, band mix, and compensation programs (including the Salary Program, sign-on bonuses, and cash retention). These rules validate approximately 18 million records daily, reducing data quality analysis time from days to minutes, which builds trust and enables leadership to make proactive, data-driven decisions.
• Data Uniformity: We ensured that all systems consuming people data—whether for reports, analytics, chatbots, or other applications — draw from either the system of record (HRMS, ATS, LMS) or Workforce360, IBM's people data platform. This approach minimizes discrepancies, creating a unified data environment.

At IBM, we recognize that high-quality people data goes beyond metrics—it’s about building trust, driving focused actions, and enabling AI to deliver meaningful recommendations. By empowering users, standardizing reports and governance, and automating data integrity validation, we've ensured that our data quality directly enhances business operations.

A Note on AI: When it comes to AI, additional quality mechanisms are essential. While strategies like bias and ethical mitigation, model fine-tuning, and hallucination detection are crucial, they can only be effective if the underlying data is of high quality. Otherwise, it’s a classic case of “garbage in, garbage out” (GIGO).

Thank you for taking the time to read this article. If you're interested in these topics and IBM’s internal people data transformation, I would be delighted if you joined my People Data Platform monthly newsletter.