Mastering data governance for effective people data platforms: lessons from what we did at IBM

This is the fifth chapter of our people data platform newsletter, examining how IBM transformed its internal platform to meet the growing demands for people analytics.

In this article, we delve into a topic that is often overlooked or undervalued: Data Governance. Designing a data platform with a primary focus on technology alone poses significant risks. Neglecting to integrate a robust data governance framework from the start can lead to serious issues later, especially regarding data quality and trust.

While there are various definitions of governance, I personally prefer a very simple one:

Data Governance is the process that ensures the availability, usability, integrity, and security of data in enterprise systems.

Within the scope of governance, we will focus on three key elements that we have found to be most crucial for our platform transformation: trust, transparency, and compliance.

Trust

People data, previously confined into silos and accessible only by people analytics experts, is now broadly accessible. Business users within and outside HR can self-serve insights, and data scientists to develop AI solutions at scale. This democratization of data highlights the critical need for users trust when using this information for talent-related decisions. Trust is supported by three fundamental components:

• Consistency: Emphasize use-case-specific standard definitions rather than universal ones. Accept different definitions for distinct purposes (e.g., headcount for talent vs. financial decisions) as long as they are clear and documented. Understand the real need for differentiation to avoid unreliable insights from inconsistent metrics.
• Continuous monitoring & communication: Maintain data integrity by regularly monitoring and promptly addressing anomalies. Build user trust and the platform's credibility by spotting and communicating issues before your users discover them, ensuring data reliability.
• Quality: Focus on ensuring data accuracy relevant to the specific use-case, recognizing that perfect accuracy is often unfeasible. Implement data governance and validation to manage variability, assessing whether the accuracy level suffices for the decision-makers' needs. For example, in attrition analysis, trend identification usually outweighs the need for exact values.

Transparency

Ensure clarity on the methods of data collection, storage, usage, and sharing, making certain that all stakeholders are well-informed about data management practices and policies.

• Data Literacy: Cultivate an understanding of data, metrics, and AI processes across the organization, highlighting their implications. Simplify access to learning resources and emphasize specialized training to enhance skills. 
• Change Management: Develop a systematic approach to change management. When modifications to data metrics are proposed, or there's a need to retire or alter data, identifying the decision-makers and stakeholders to be informed is crucial. Ensure that once decisions are reached, they are communicated efficiently through the enterprise. Avoid bypassing the process and perhaps adding new data solely to meet temporary needs, as this decision may end up haunting you indefinitely.
• Data Lineage: Monitor and effectively communicate the origins and development of metrics and analytics. Keep in mind the varying levels of people analytics expertise throughout the organization, and avoid assuming universal familiarity with these concepts, even if they were created years ago.

Compliance

In recent years, the management of people data has significantly increased in complexity due to the emergence of new privacy laws worldwide and more stringent expectations that companies have rightly established for the ethical use of advanced AI solutions. Three aspects:

• Regulation compliance: It is essential to adhere to relevant laws, regulations, and standards governing data protection, privacy, and security.
• Ethical usage: Data should be used in ways that respect individual rights and societal norms, avoiding manipulation, discrimination, or any form of misuse.
• Swift adapt to change: The management of people data is constantly evolving, often necessitating immediate changes in data management to align with corporate directions. Failure to apply such changes at speed and across all processes using people data can result in costly legal and regulatory issues.

Effective data governance is an excellent way to secure executive support and sponsorship for a data platform, highlighting benefits such as reduced compliance risks, and better, more reliable talent insights. This approach benefits not only HR but the entire organization.

Our experience with Workforce360

At IBM, we embraced "governance-by-design" as a fundamental principle in the implementation of Workforce360. This approach involved resisting the temptation to take shortcuts and instead, concentrating on developing a solution that leverages enduring technology.

Trust

At WF360, our obsession with data and metric consistency drives us to only ingest data that meets three key criteria: (A) a committed data owner, (B) a clear use-case (with quantifiable business impact), and (C) thorough testing to ensure its reliability over time. Defining or altering people metrics (e.g., Attrition and Headcount) involves a rigorous change-management process and two-tiered governance. Within HR, every metric is linked to an owner and a list of decision-makers and Subject Matter Experts (SMEs). Decision-makers can formally object to changes, but they need to provide (with the help of SMEs) a firm business impact justification. Once approved by HR, the data undergoes a second-tier governance process at the enterprise level, involving data steward representatives from outside HR.

In terms of data quality, we utilize IBM Knowledge Catalog. Its automation and anomaly detection capabilities help us maintaining high data standards and provide reliable talent insights. Currently, we have a few dozen rules in production that address various aspects of HR workflows, encompassing organizational hierarchy, attrition, band mix, and most notably, compensation programs, which include the Salary Program, sign-on bonuses, and cash retention. When these rules are executed, they validate roughly 18 million records. Data quality analysis has been significantly expedited, reduced from days to just minutes. This speed enables us to proactively and consistently alert the appropriate user when specific data changes need their attention, thereby enhancing trust.

Transparency

At IBM, we direct HR and non-HR professionals interested in utilizing data to "Your Insight," a user-friendly companion to Wf360 designed for non-technical users. "Your Insight" serves as a comprehensive resource for reports, analytics, data, support, and particularly data governance.

Our data catalog contains key details of reach piece of data ingested in the platform, including the name and description of the data, its source (such as where WF360 gets the data), how often it is ingested (e.g., daily), and guidelines on how to access it. By providing a reliable, unified source of information, we've greatly minimized the time spent discovering data, and reduce the risk of errors and inconsistencies, thereby enabling users to make more informed decisions.

People data metrics are presented in alignment with the specific business needs they serve, recognizing that different definitions for the same metric may be appropriate. For example, in external reporting, we might estimate the number of equivalent full-time employees considering temporary, part-time, and limited-term employment. Conversely, for metrics like Learning completion, we make no distinctions based on employment type. We found transparently selecting the correct metric for each use-case is even more crucial than selecting the correct data for it. 

Compliance 

Over the years, Workforce360 has grown by incorporating significant amounts of Personally Identifiable and Sensitive Personal Information (PI/SPI) from IBM employees across 170 countries. Our commitment to compliance is underpinned by a strategy of "standardization." We streamline our operations by standardizing our data, eliminating more than a dozen redundant copies, standardizing technology through phasing out old systems, and working closely with the Enterprise Data team. We also standardize access processes to ensure they are both comprehensive and adaptable. This approach allows us to quickly modify access rights in line with specific country regulations or internal policy shifts. In the next article of this series, I will delve deeper into our approach for granting data access. 

Ethics are a cornerstone of IBM’s operations. The AI Ethics Board, a diverse group from various sectors of IBM, plays a pivotal role in embedding our ethical principles into any AI solutions built on Workforce360. Additionally, IBM enforces tech responsibility by adhering to the IBM Principles for Trust and Transparency and our pillars for trustworthy AI.

More details on how IBM manages the ethical use of AI can be found here. At IBM, while we trust our employees to act correctly, we acknowledge that not everyone, particularly those outside HR, may be fully equipped in understanding the nuances of people data. Thus, the people data platform is critical in maintaining compliance. When AI data requests are made, our team reviews all necessary authorizations and guides the requester on proper usage. 

This article highlighted the critical role of data governance in developing a data platform. We explored three key considerations: Trust, Transparency, and Compliance. These factors are particularly important in managing people data, for two main reasons: (A) As the people data platform encourages democratization, it's important to recognize that users may not know how to effectively utilize such data for insights. (B) People data must adhere to strict regulatory requirements, which vary globally.

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.