Most Fabric engagements fail the same three ways: platform before purpose, underestimated data quality, and no capability transfer. What a good engagement covers, and the questions to ask a partner.
Most Microsoft Fabric engagements fail in the same three ways. They start with the platform instead of the question, they underestimate the data quality work, and they leave the agency unable to run the thing once the consultants leave. Here is what a Fabric engagement should actually look like.
Start with the decision, not the architecture
A good engagement opens by asking which decisions the agency currently makes badly or slowly because the data is not available. That produces a shortlist of concrete outputs. The architecture then follows from those outputs. The reverse order, where a reference architecture is drawn first and use cases are retrofitted, reliably produces an expensive platform that nobody uses.
Budget honestly for data quality
In our experience the technical build is rarely the hard part. The hard part is discovering that three systems hold three different versions of the same entity and that nobody owns the reconciliation. A realistic Fabric programme allocates meaningful effort to profiling source data before pipelines are built, and it names a business owner for each domain. If a proposal does not include that, it is understating the cost.
Size the capacity with evidence
Fabric capacity is a reserved commitment. Oversize it and you burn budget on idle compute. Undersize it and queries throttle at month end when everyone runs reports at once. The only reliable way to get this right is to build one representative workload first and measure it. Treat any capacity recommendation made before that measurement as a guess.
Sovereignty and data residency
For Australian government workloads, confirm early which Fabric services are available in Australian regions and which are not, because the gap moves. Confirm where Copilot and other AI features process data. Confirm what your agency's classification framework permits. These answers shape the design and are far cheaper to establish at the start than after a build.
Questions worth asking a partner
- How will you size our capacity, and at what point in the engagement?
- What proportion of the effort is data quality and governance rather than build?
- Which Fabric services are you proposing that are not yet generally available in Australian regions?
- How does our team take this over, and what does the handover actually consist of?
- Can you show a workload you built that the client now runs without you?
Plan for the handover from day one
The measure of a good Fabric engagement is whether the agency can extend the platform six months later without calling anyone. That means the internal team builds alongside the partner rather than watching, naming conventions and patterns are documented as they are set rather than at the end, and the operating model is agreed before go live. Capability transfer is not a phase at the end of a project. It is the point of the project.