Three Member Counts, One Council: Mandating an Enterprise Metric
Northbridge Health, a regional insurer with four business units, has run a data governance council for eight months. Its catalog holds 4,200 assets, but 38% have no assigned owner and only 11% have documented definitions. The immediate trigger is "active member count", which appears on 11 executive dashboards with three different values (2.41M, 2.37M, 2.29M) because claims, enrollment, and finance each apply different inclusion rules: enrollment counts members inside a 60-day grace period, finance excludes them as not yet premium-paying. The CDO wants to mandate a single enterprise definition and freeze dashboard changes until reconciliation completes. Business unit leads resist: the regulatory filing uses the finance scope and cannot change mid-year, and two units argue their operational decisions genuinely need the enrollment view. A previous "big bang" glossary effort two years ago produced 900 definitions that nobody used, and stewards still cite it as a reason to move slowly. Three stewards, no additional headcount.
1. The CDO asks whether to mandate one enterprise definition of active member count and freeze dashboards. Evaluate the options and recommend a course of action, including the governance artifacts required.
Read the worked answer
The mandate as framed conflates two problems: three different numbers, and three legitimate scopes. Start by decomposing the 120k spread between 2.41M and 2.29M against a single as-of date - reconcile record by record into grace-period members, retroactive terminations, and duplicate enrollments. That tells you whether this is a definition dispute or a data defect; often it is 80% defect, 20% scope.
Recommend a certified metric with registered scope variants rather than one number. One calculation base (source of record, grain, as-of semantics, standard exclusions), plus three approved variants - regulatory, finance, operational - each with an owner, purpose, and validity window. The regulatory variant is bound by filing rules and cannot be overridden by the council; that is a constraint, not a negotiation.
A blanket dashboard freeze is disproportionate and will be routed around by teams building shadow queries. Instead, require every existing dashboard to display which variant it uses, and require new dashboards to consume the certified metric only.
Artifacts: a metric definition record with calculation logic and lineage; a variant register with owners and approval dates; certification status per dashboard; a deprecation schedule for legacy queries; a change log tied to the filing calendar. Decision rights: metric owner proposes, council approves base logic, DPO/compliance signs off on the regulatory variant. Success is measured by the number of bespoke member-count queries retired, not by glossary entries created.
2. With three stewards, no budget, 38% of 4,200 assets unowned and 89% undefined, how should ownership and definition work be sequenced over the next two quarters?
Read the worked answer
You cannot cover 4,200 assets with three people, and attempting it produces exactly the 900-definition artifact nobody uses. Work backward from consumption instead of forward from the catalog.
Quarter 1: identify critical data elements. Rank by (a) regulatory or statutory reporting obligations, (b) revenue and billing impact, (c) executive and board metrics, (d) asset usage from query logs and lineage. A realistic outcome is roughly 300 Tier-1 assets and 40–60 critical data elements. Assign ownership at domain level - eight domain owners, not 300 asset owners - and require each to name a steward for Tier-1 elements only. Set a hard target: 100% of Tier-1 elements have an accountable owner and a one-page definition by end of Q1. Tier-2 and Tier-3 assets get automated technical metadata only.
Quarter 2: definitions for the top 40 metrics that drive decisions, each with calculation logic, source, grain, and known variants. Reuse templates; reject definitions longer than a page. Use lineage and query logs to prove reuse - if a definition has no consumer, it is not done, it is decoration.
Report leading indicators, not coverage: percentage of executive dashboards on certified metrics, median time to resolve a metric dispute, percentage of Tier-1 elements with an owner, and steward hours per defined asset. Coverage percentage is a vanity metric that rewards documenting trivial columns.
3. Two business units insist their operational view must stay on their own dashboard. How do you keep the certified metric from becoming a fourth conflicting number, and how do you know governance is working?
Read the worked answer
The risk is real: if the certified metric is published but operational teams keep their own variant unlabeled, you have added a fourth number and lost credibility. Treat the operational variant as first-class - register it, name its owner, state its purpose and its known divergence from the finance base (grace-period members, expected magnitude, direction), and require the dashboard to carry that label next to the figure.
The test is explainability: any two numbers on any two dashboards should be reconcilable to a documented scope difference in under a day. If a business unit wants a new variant, it goes through a lightweight request with a stated decision it supports; the council approves or rejects on that basis, not on politics.
Measure whether governance is working with: number of unregistered member-count queries found in audit; percentage of dashboards displaying variant labels; reconciliation time between finance and enrollment figures (target under one business day, from the current multi-week cycle); number of variant requests rejected versus approved; and repeat disputes on the same metric. A drop in repeat disputes is the strongest signal. Counting glossary terms or catalog assets will tell you nothing about whether anyone trusts the number.