340 Critical Data Elements, Six Part-Time Stewards
Metro Health System appointed its first CDO seven months ago. A census identified 340 critical data elements across patient, provider, and finance domains. Only 22% have a named accountable owner. Six data stewards work 20% on governance alongside operational roles; the rest of the organization treats requests as an audit tax. The governance council has met twice, and 140 data quality incidents are open, 31 affecting regulatory submissions. Internal audit cited inconsistent customer due diligence data across three systems. The CEO wants AI-ready data within two quarters and has asked the CDO for a plan that does not add headcount. Business unit leaders say they will support governance once they see value, and the CIO wants a single enterprise data catalog purchased this year.
1. How would you prioritize the 340 critical data elements given six part-time stewards and two quarters?
Read the worked answer
Do not launch a 340-element program with six part-time stewards; it will fail visibly. Tier the elements by consequence: tier 1 is the 25-30 elements tied to the audit finding, regulatory submissions, and the first AI use case; tier 2 is elements with clear financial or customer impact; tier 3 is the rest, documented but not actively stewarded this year. For tier 1, assign accountable owners at the business-unit VP level, not the steward level, and define a measurable acceptance criterion for each. Use an impact-versus-effort matrix and a RACI that distinguishes accountable, responsible, consulted, and informed. Reuse existing forums such as the revenue cycle or quality committee rather than creating a new one. The AI-ready goal should be scoped to a named use case with known data dependencies, not treated as enterprise-wide certification. Publish the tiering rationale so business units see why some requests wait. This is a portfolio decision under a hard capacity constraint.
2. Recommend an operating model and decision rights that fit a no-new-headcount mandate.
Read the worked answer
Adopt a federated model with central coordination, because the CDO has no headcount to centralize. The central team owns standards, the catalog, the CDE register, and escalation; business units own their data and name stewards. Convert the six part-time stewards into domain leads with explicit role descriptions and time commitments approved by their managers, and recruit additional stewards from domains with tier-1 elements. Decision rights must be written: business owners decide definitions and accept risk; the CDO challenges and maintains standards; IT implements; internal audit provides independent assurance. The governance council should meet monthly with a fixed agenda of exceptions, not status updates, and have a quorum that can decide. Create an escalation path with service levels for unresolved cross-domain conflicts. Avoid a paper RACI; test it on three live incidents. The judgment is that governance without decision rights is just meetings, and volunteer stewards without managerial backing will deprioritize governance when operations spike.
3. What metrics would you use to prove value and sustain funding?
Read the worked answer
Use a balanced set of leading and lagging measures, and do not count meetings or catalog entries as value. Lagging: audit findings closed, regulatory submission defects, incident mean time to resolve, duplicate or error rates on tier-1 elements, and cycle time to onboard a new AI data source. Leading: percentage of tier-1 elements with named accountable owners, percentage with documented definitions, steward capacity versus assigned elements, and percentage of incidents resolved within SLA. Add adoption and trust measures such as the share of AI features using certified data and business self-service usage of the catalog. Report quarterly with a baseline and a small number of targets; too many metrics dilute focus. To win over skeptical business units, publish two or three concrete wins, such as a cleared audit finding or a faster regulatory submission, with before-and-after numbers. The judgment is to measure outcomes the CEO and CFO already care about, and to resist optimizing for coverage of all 340 elements.