Skip to main content

Power BI Governance Center of Excellence — operating model, tenant controls, certified semantic models, deployment pipelines

Last updated by Errin O'Connor, Founder & Chief AI Architect, EPC Group

EPC Group builds the Power BI governance Center of Excellence in three stages: a three-week assessment that scores six pillars — operating model, tenant controls, certified semantic models, deployment pipelines, capacity and cost, adoption — with evidence and a ranked findings register; a remediation stage that decides the tenant settings, certifies the first subject areas, stands up deployment pipelines, retires duplicate content and right-sizes capacity against Microsoft list pricing; and the installation of the operating model with your people in the roles. Fourteen weeks, a senior architect end to end, 1,500+ Power BI deployments behind the method. Fixed-scope, priced after a scoping call.

Key Facts

  • Six pillars scored 0–3 with evidence; findings ranked as blockers, maturity gaps or observations with owners
  • Three stages: assessment (weeks 1–3), remediation (weeks 4–10), Center of Excellence installed (weeks 11–14); optional run stage
  • One certified semantic model per subject area — the control that makes the others matter, and surface 5 of TAR-8
  • Tenant controls decided setting by setting with the security group and the reason recorded
  • Deployment pipelines or Git integration replace publishing from laptops to production
  • 1,500+ Power BI deployments · 500+ Fabric implementations · Microsoft consulting since 1997
  • Capacity plan against Microsoft list pricing; contracted records in the Evidence Center

Why Power BI governance is an operating model, not a settings page

Every Power BI tenant has settings. Very few have an owner for them. Governance fails not because the controls are missing from the product but because nobody decided them, nobody certified the models the business relies on, nobody stopped reports being published from laptops to production, and nobody watched the capacity until it throttled at month-end. Then Copilot arrives and answers from the fiftieth version of revenue. The fix is a function — people in roles, a board that decides, a process that certifies and releases — installed on top of controls that were set on purpose. That is what a Center of Excellence is, and it is what this engagement builds.

The six pillars

1

Operating model

Who owns the tenant, the capacities, the certified models and the workspaces; the governance board and its cadence; the roles (platform owner, data steward, model owner, report author, consumer) and what each may do; the intake and review process for new workspaces and models; the metrics the board watches.

2

Tenant controls

The Power BI admin settings decided deliberately rather than left at default — export and sharing, external users, publish-to-web, custom visuals, sensitivity-label enforcement, data-loss prevention, the developer and Fabric settings — scoped to security groups and documented with the reason for each.

3

Certified semantic models

One certified model per subject area as the single source the business and Copilot answer from; row-level and object-level security; the endorsement process; the retirement of duplicate and personal-workspace models; Direct Lake versus import versus DirectQuery decided on evidence.

4

Deployment pipelines and ALM

Development, test and production workspaces; deployment pipelines or Git integration; release rules and approvals; parameterized data sources and gateways; the end of publishing straight from a laptop to production.

5

Capacity and cost

Premium or Fabric capacity sized and allocated against utilization, refresh schedules and workloads tuned, the workloads that burn capacity for little value identified, and the cost plan written against Microsoft list pricing.

6

Adoption and enablement

Author training tied to the standards, a report catalog, usage metrics, a retirement policy for unused content, and a Center of Excellence community that answers questions so the platform team does not have to.

From assessment to remediation to a team that runs it

Assessment (weeks 1–3)

The six pillars scored 0–3 with evidence — admin-portal exports, the capacity metrics app, the activity log, the model inventory — and every finding classified as a blocker, a maturity gap or an observation with an owner and a closing control.

Remediation (weeks 4–10)

The tenant settings decided and applied; the certified-model standard set and the first subject areas certified; deployment pipelines stood up and the first reports moved through them; duplicate and orphaned content retired in waves; capacity right-sized.

Center of Excellence (weeks 11–14)

The operating model installed with your people in the roles, the governance board running, the intake process live, the catalog and metrics in place, and the community launched. EPC Group hands over a function, not a document.

Run (optional)

The same architects operate the function under Power BI admin and governance support — tenant and capacity administration, model governance, security posture monitoring and incident response — until your team is ready to take it, or indefinitely.

What the assessment delivers on its own

The three-week assessment is complete without the later stages: the six-pillar scorecard with its evidence column; the findings register with classification, owner and closing control; the tenant-settings decision sheet; the semantic-model inventory with the certification candidates; the capacity utilization analysis with the right-sizing recommendation against Microsoft list pricing; and the roadmap, delivered with a 90-minute executive readout. A redacted sample of the set is walked through on the scoping call — the shape and depth are what you are judging; its numbers belong to that client and are not published.

Why EPC Group

Frequently Asked Questions

What is a Power BI governance Center of Excellence?

The function that keeps a Power BI estate trustworthy at scale: an operating model with named owners and a governance board, tenant controls set on purpose, one certified semantic model per subject area, deployment pipelines between development, test and production, capacity managed against cost, and the adoption program that makes authors follow the standards. EPC Group builds it in three stages — assessment, remediation, Center of Excellence — on 1,500+ Power BI deployments and 500+ Microsoft Fabric implementations.

Who is it for?

Organizations whose Power BI grew faster than their governance: hundreds of workspaces, dozens of versions of the same metric, reports published from laptops to production, tenant settings still at default, a capacity that throttles at month-end, and now a Copilot or Fabric data-agent program that will answer from whatever is there. Also organizations moving from Power BI Premium to Fabric who want the governance built before the migration, not after.

How does the assessment work?

Three weeks. The six pillars — operating model, tenant controls, certified semantic models, deployment pipelines, capacity and cost, adoption — are scored 0–3 with the evidence behind each score, and every finding is classified as a blocker (fix before the estate grows or before AI is enabled on it), a maturity gap (fix in the first quarter) or an observation, with an owner and the Power BI or Fabric control that closes it. The assessment can stand alone; most clients continue into remediation.

What does remediation change?

The tenant settings are decided and applied with the reason for each documented; the certified-model standard is set and the first subject areas certified with row-level security; deployment pipelines or Git integration are stood up and the first reports released through them; duplicate, orphaned and personal-workspace content is retired in waves; capacity is right-sized against utilization and the cost plan is written against Microsoft list pricing.

What does EPC Group leave behind?

A function, not a document: your people in the roles, the governance board meeting on a cadence, the intake and certification process running, the report catalog and usage metrics live, the community launched, and the standards written for the people who will apply them. The optional run stage keeps the same architects operating it under Power BI admin and governance support.

Why are certified semantic models the center of it?

Because they are what the business, Copilot and Fabric data agents answer from. A tenant with fifty uncertified versions of revenue has no governance whatever its settings say. One certified model per subject area, with row-level security and an endorsement process, is the control that makes every other control matter — and it is surface 5 of the TAR-8 tenant AI readiness standard.

How do tenant controls get decided?

Setting by setting, with the security group that scopes it and the reason recorded: export and sharing, external users, publish-to-web, custom visuals, sensitivity-label enforcement, DLP, developer settings, Fabric settings. Defaults are a decision someone else made; the engagement replaces them with decisions your governance board owns.

How does this differ from Power BI admin and governance support?

Admin and governance support is the run stage — EPC Group operating the function on an ongoing basis. The Center of Excellence engagement is the build: assessment, remediation and the installation of the operating model with your team in the roles. Many clients do the build, then the run; some do the build and take it over.

How does it relate to the Fabric governance assessment?

The Fabric governance assessment scores the whole Fabric estate — OneLake, domains, workspaces, capacity, Purview — and is the right start when the platform is Fabric-first. This engagement goes deep on the Power BI layer and builds the function that governs it. On a Fabric estate the two are sequenced together; on a Power BI Premium estate this one stands alone.

How long does it take, and what do you need from us?

Fourteen weeks for a single-tenant estate from assessment to a running Center of Excellence, scoped up front; the assessment alone is three weeks. We need an executive sponsor, the Power BI or Fabric administrator, the data-platform lead, the finance owner of the capacity budget, and the people who will hold the CoE roles from week eleven. Multi-tenant or multi-geo estates are scoped in writing before work starts.

How is it priced?

Fixed-scope, priced after a scoping call that confirms the number of workspaces, models and capacities in scope and whether the assessment, the full build or the run stage is wanted; there is no rate card. Power BI and Fabric licensing and capacity are quoted at Microsoft list price. Contracted engagement records for comparable analytics governance work are published, redacted, in the EPC Group Evidence Center.

Related EPC Group services and references

Book the scoping call

Thirty minutes with the architect who will lead the engagement. Bring your workspace count and your capacity metrics; you leave with the scope, the participant list and the redacted sample walkthrough.

Book the scoping call Call (888) 381-9725

Related reading

AI assistant — not human