The Power BI Fellowship
Patterns & Playbooks · Guide

Governance and the BI lead

Workspace strategy, naming, ownership, certification, stewardship, content lifecycle, retirement, self-service vs managed vs enterprise BI, the Center of Excellence and governance maturity, built on Microsoft's Fabric adoption roadmap.

Senior BI DeveloperBI Architect / LeadPower BI ServiceMicrosoft FabricPower BI REST APIChecked 2 Oct 2026Download .md
On this page (11)

What governance is for

Governance isn't a 40-page framework. It's the smallest set of roles, rules and habits that lets people trust the numbers, find the right content, and build their own reports without creating chaos. Microsoft's Fabric adoption roadmap is the authoritative foundation; this page is the practical summary.

Rule: Make the governed path the easiest path. If the certified model is harder to use than exporting to Excel, people will export to Excel.

Workspace strategy

PatternWhen
By domain and stage: Sales Analytics [Dev], [Test], prodContent managed by a BI team with deployment pipelines
Data workspaces separate from report workspacesShared semantic models consumed by many report authors
Personal and team sandboxesSelf-service exploration, with no expectation of support

Group workspaces into domains (Sales, Finance, Operations) so ownership and discovery follow the business.

Naming conventions

Consistent names make inventories, lineage and scripts work. Agree patterns for workspaces, semantic models, reports, apps and security groups, and publish them. Download the workspace naming standard.

Ownership models

ModelWho buildsWho owns data definitions
Business-led self-serviceBusiness analystsThe business unit
Managed self-serviceBI team builds certified models; business builds reports on themBI team with business data owners
Enterprise BICentral BI team builds models and reportsCentral team with data owners

Most organisations mix all three. Write down who owns each certified model and each key definition (Revenue, Customer, Headcount) in an ownership matrix.

Semantic model certification

Write the bar for certification down and apply it every time: owner, descriptions, tests, refresh monitoring, RLS tested, documented definitions. Use the certification checklist.

Data stewardship

A data steward for each domain answers "what does this field mean?" and "is this number right?", approves definition changes, and is the person a developer asks before changing a KPI. Without stewards, every disagreement escalates to the BI team.

Content lifecycle and report retirement

Content has a lifecycle: create, publish, promote or certify, maintain, retire. Retirement is the step everyone skips:

  1. Find candidates: no views in 90 days, superseded by certified content, no owner (find stale content).
  2. Contact owners and recent viewers; many "unused" reports are used quarterly.
  3. Announce the date and the replacement.
  4. Remove access or move to an archive workspace for a grace period.
  5. Delete after the grace period, keeping a backup (PBIP in Git, or an export).

The deprecation checklist and the report retirement drill walk through it.

Self-service, managed self-service and enterprise BI

Self-service is a feature, not a failure. The goal is to channel it: give authors certified models (Build permission), training and templates, and keep enterprise BI for the content that must be right for everyone. Monitor with usage metrics and the activity log, not with bans.

Center of Excellence

A COE is the small team that makes BI succeed across the organisation. Microsoft's guidance describes typical COE responsibilities including training and enablement, documentation and templates, governance, user support, architecture and platform oversight. A COE doesn't build everything; it makes everyone else better at building. Start with a COE charter that names its mission, scope, people and the few metrics it will report.

Governance maturity

Assess where you are before planning where to go. The roadmap's maturity levels run from initial (no consistent practice) through repeatable and defined to capable and efficient. Use the governance assessment and practise in the governance maturity drill.

The BI lead's first quarter

  1. Inventory what exists (workspaces, models, reports, owners, usage): Sprint 10 is this exercise.
  2. Pick the three definitions that cause the most arguments and certify one model for them.
  3. Name owners for every certified model and every workspace with production content.
  4. Publish the few rules (naming, certification bar, where to ask for help) on one page.
  5. Retire the obvious dead content, announced and reversible.
  6. Report progress with numbers: certified-model usage up, duplicate models down, incidents down.

Something missing or out of date? Open an issue or edit content/toolkit/guides/governance-lead.md.