Confidential engineering organization · anonymized case study
Catching Fabric migration risks before production
How a Microsoft Fabric architecture assessment turned hidden governance, ingestion-cost, and Direct Lake risks into a sequenced remediation roadmap.
This is an anonymized engagement. The client name, environment identifiers, proprietary workloads, and internal commercial details are intentionally withheld.
Client context
The situation.
An engineering organization was migrating a legacy Synapse environment into a Medallion Architecture on Microsoft Fabric. The team had a thoughtful semantic framework and real CI/CD discipline, but the physical Lakehouse design carried risks that would emerge at production scale.
Symptoms and risks
- More than 100 tables in one shared schema
- Destructive full-table-overwrite ingestion patterns
- Capacity Unit exposure hidden inside workload design
- A Gold layer not ready for Power BI Direct Lake
Review approach
How the problem was examined.
The work focused on evidence and decision quality before prescribing implementation.
Reviewed workspace and access governance
Assessed physical Lakehouse organization
Examined ingestion and workload patterns
Evaluated consumption-layer and Direct Lake readiness
Engagement timeline
How the work unfolded.
Phases are listed in the order they happened. Durations appear only where they can be stated without exposing client detail.
- 01
Governance review
Reviewed workspace boundaries, access patterns, and who could reach what once the environment carried production data.
Access exposure surfaced before it became an incident.
- 02
Lakehouse organization assessment
Assessed the physical layout: more than 100 tables sharing one schema with no isolation between domains or lifecycle stages.
- 03
Ingestion and workload patterns
Examined the full-table-overwrite ingestion pattern and where Capacity Unit consumption was hiding inside workload design.
Compute-budget exhaustion identified as a scale risk, not a tuning problem.
- 04
Consumption-layer readiness
Evaluated whether the Gold layer could serve Power BI Direct Lake as designed, including Star Schema and V-Order requirements.
- 05
Roadmap handoff
Delivered a three-phase sequence: schema-level isolation, metadata-driven CDC ingestion, then Direct Lake preparation. Implementation stayed with the client's platform team.
A prioritized, approachable roadmap rather than an overwhelming list of improvements.
Findings and recommendations
A sequenced path, not an unbounded backlog.
- Phase 1: establish schema-level isolation
- Phase 2: replace full overwrites with metadata-driven CDC ingestion
- Phase 3: prepare Star Schema and V-Order patterns for Direct Lake
Outcome
What changed.
The client received a concrete, prioritized sequence for closing the gaps before they became production incidents, including access exposure, compute-budget exhaustion, and consumption-layer limitations.
Client perspective
What the client said about the assessment.
A public reference from the organization this engagement served. The sector line above stays anonymized on purpose.
Working with Chris was a great experience. He has a very approachable, conversational style that made it easy to discuss complex topics without getting overwhelming. Throughout the engagement he balanced technical depth and best-practice with practicality and feasibility.
What we appreciated most was that he came into the assessment with a clear structure and left us with a prioritized, approachable roadmap rather than an overwhelming list of improvements. He helped validate many of the directions we were considering, identified areas we could avoid future tech debt, and provided practical recommendations we can realistically execute over time.
I would absolutely recommend Chris to any team looking for a thoughtful technical advisor who can communicate effectively with both engineers and leadership.
Related service
Explore data architecture consulting
Review the service boundaries, typical deliverables, pricing context, and evidence before deciding whether to schedule.
Your situation
Facing a version of this problem?
Bring the platform, governance, reliability, or migration issue that is difficult to explain or expensive to keep delaying. Thirty minutes to work out whether it is a fit.