Governance, Catalog, and Self-Service

This topic evaluates how BI assets are organized: who owns what, which models are certified and reused, how workspaces are structured, how copy sprawl is avoided, and how…

Guide≈ 6 min readCheckedReading

In brief

This topic evaluates how BI assets are organized: who owns what, which models are certified and reused, how workspaces are structured, how copy sprawl is avoided, and how self-service is governed (analysis from shared models, composite models) without killing it. Enterprise data governance (global catalog, stewardship) is in the PD matrix; here, we cover governance of BI artifacts.

On this page

1. Why this topic
2. What the topic measures
3. The 5 levels
4. Milestones by level
5. Detailed reading by level
6. Partial achievement and rating
7. Moving to the next level
8. See also

Level: 🟢→🔴 · Theme code: GOV

1. Why this topic#

Without governance, a BI tool's success creates its own debt: everyone duplicates the model “just to be safe,” five versions of revenue coexist, no one knows which report is authoritative, and old reports keep consuming capacity. With governance that is too heavy, business teams work around it with Excel. The goal is a lane: wide enough for self-service to move forward, structured enough to keep a single source of truth.

2. What the topic measures#

Axis: from file proliferation to a living catalog of BI products, with safe self-service.
Key indicators: share of reports connected to a certified model; number of duplicated models; share of artifacts with an owner and review date; ratio of viewed reports/reports that exist.
Units: %, ratio.

3. The 5 levels#

Level	Short name	Promise	Typical indicator (order of magnitude)
1	Proliferation	Everyone finds or builds their own report; no one knows which one is authoritative.	No inventory; duplicates
2	Known owners	Each artifact has an identified owner and workspace.	100% of reports with an owner
3	Certified models	Official numbers come from a recognized model that everyone can identify.	≥ 60% of reports on a certified model
4	Living catalog	Artifacts can be found, assessed, and retired according to an explicit life cycle.	Annual review of each artifact; retirement of unused ones
5	Safe self-service	Business teams build their analyses on official models without copying them or bypassing rules.	≥ 90% “thin” reports; 0 competing official number

4. Milestones by level#

ID	Milestone	Verifiable evidence
BI-GOV-1.1	A list (even imperfect) of production reports exists	Inventory
BI-GOV-1.2	Personal workspaces are not used for production	Control of prod reports
BI-GOV-1.3	A contact point for BI requests exists	Official channel
BI-GOV-2.1	Each report/model has an owner and a backup	Owner field
BI-GOV-2.2	Workspaces follow a convention (by domain and environment)	Naming rule applied
BI-GOV-2.3	Workspace permissions follow groups (RLS)	Permissions review
BI-GOV-2.4	Description and purpose on every published artifact	Descriptions filled in
BI-GOV-3.1	“Certified/promoted” models (endorsement) for official numbers	Labels in the service
BI-GOV-3.2	Written certification criteria (tests, documentation, owner, SLA)	Criteria document
BI-GOV-3.3	Official reports connect to the certified model (no copy)	Share of thin reports
BI-GOV-3.4	Key business definitions (active customer, completed visit) are published	Business glossary
BI-GOV-3.5	Dependency links (report → model → source) can be viewed	Viewable lineage
BI-GOV-4.1	Life cycle: creation, certification, review, deprecation, retirement	Procedure applied
BI-GOV-4.2	Catalog in use (search, labels, domains)	Up-to-date catalog
BI-GOV-4.3	Regular usage review: unused reports are flagged then retired	Retirement log
BI-GOV-4.4	Change management: model changes are announced to consumers	Announcement log
BI-GOV-5.1	Guardrailed self-service zone: certified models + “sandbox” space	Dedicated spaces
BI-GOV-5.2	Promotion of a self-service report to official through a defined path	Documented path
BI-GOV-5.3	Automated governance checks (ownerless reports, risky sharing)	Automated reports
BI-GOV-5.4	Community of practice and continuous training	Sessions held

5. Detailed reading by level#

🟢 Level 1 — Proliferation#

Signs: “Paul's report,” five variants of “Sales,” exports saved in emails.
Difficulty: 1/5 · Effort: 1/5.

🟢 Level 2 — Known owners#

Signs: inventory table, one owner per model and report, spaces named by domain/environment.
Benefit: you know who to contact.
Limit: having an owner says nothing about quality.
Difficulty: 2/5 · Effort: 2/5.

🟡 Level 3 — Certified models#

Signs: endorsement (“promoted,” “certified”) identifies official models; certification criteria are written: tests in place, RLS verified, documentation, owner, support. Official reports are thin reports tied to this model.
Benefit: one version of the truth, lower capacity cost (CST).
Limit: the certified model becomes a contention point if its delivery cycle is slow.
Difficulty: 3/5 · Effort: 3/5.

🧪 Example

SFA common thread: the certified “Visits & Sales” model feeds 14 thin reports. A regional director wants to add an indicator: they create a local measure in their thin report (allowed); if others use it, a promotion path moves it up into the certified model.

⚠️ Pitfall

Symptom: the certified model is connected to 14 reports, but each one extends it with its own local measures that redefine “active customer.” Cause: the official definition is not accessible or flexible enough. Fix: publish definitions, open a fast request channel, measure competing definitions.

🟡 Level 4 — Living catalog#

Signs: an explicit life cycle (draft → promoted → certified → deprecated → retired), annual review, retirement of unused reports (activity logs), announcements of model changes.
Benefit: cleaner assets, lower cost.
Limit: review discipline is essential.
Difficulty: 4/5 · Effort: 4/5.

💡 Expert tip

Use the admin API (inventory reports, activity logs) to automatically produce the list of artifacts without an owner, without views for 90 days, or shared by link (the activity logs API only returns the last 28 days: archive them to measure over 90 days). A manual catalog is outdated within a quarter.

🔴 Level 5 — Safe self-service#

Signs: guardrailed self-service zone (certified models + sandbox), promotion path, automated checks, community of practice. Business analysts build on official models without duplicating them. Composite models (model chaining) make it possible to add local data without copies.
Benefit: business speed, consistency of numbers.
Limit: large-scale self-service creates support demand; capacity must keep up (OPS).
Difficulty: 5/5 · Effort: 5/5.

✅ Good practice

Measure adoption of the certified model by the number of thin reports and by the disappearance of competing definitions, not by the number of people trained.

🤖 AI

Assistants (Copilot; “Q&A,” whose retirement from Power BI is announced for December 2026) rely on model descriptions, names, and synonyms: a well-documented model (MOD) gives better answers; an ambiguous model produces plausible but wrong answers. Governing what is exposed to assistants is part of this topic.

6. Partial achievement and rating#

Common rule. “A partial 5”: guardrailed self-service zone for 2 domains out of 5 (BI-GOV-5.1 = 0.5), promotion path described but never exercised (5.2 = 0.5), daily automated governance checks (5.3 = 1), informal community (5.4 = 0.5).

7. Moving to the next level#

Possible means (optional):

Transition	Possible levers
1 → 2	Inventory via admin API; owners; workspace convention
2 → 3	Certification criteria; endorsement; thin reports; business glossary
3 → 4	Life cycle; usage review; change announcements
4 → 5	Sandbox, promotion path, automated checks, community

Pitfalls:

False positive: 100% of “certified” reports because certification is granted with no criteria.
Goodhart: reducing the number of reports by deleting without checking usage; seasonal reports (year-end close) appear unused ten months out of twelve.
Control governance without service: centralization with no response time pushes workarounds.
The certified model becomes a bottleneck: plan for governed local extensions.

8. See also#

Related topics: MOD · RLS · OPS · CST · TST
Other matrices: PD — Data Governance and Organization · PD — Metadata, Lineage, and Documentation · PD — Adoption and Change Management
Content: Cross-Cutting Power BI Good Practices · Power BI Reporting Project Checklists · OneLake + SaaS Mobile Apps: Overview · Source Control for Data: SQL and Power BI

✻ Written by Claude Sonnet 5.5