Unifying a Global Business Services Firm’s Fragmented Azure Data Estate into a Production-Grade Fabric Platform

4 min read

Mar 5, 2026 1:45:26 PM

Pythian consolidated five legacy Microsoft services into a single governed lakehouse

A global business services organization running operations across 30+ countries had outgrown its patchwork of Azure Synapse, SSIS, Azure Data Factory, Power BI Premium, and ADLS Gen2. Siloed environments, inconsistent metrics, and escalating cloud costs blocked every attempt at production analytics. Pythian consolidated the entire estate into Microsoft Fabric—on time, on budget, with zero unplanned downtime.

35%

Reduction in Azure data platform spend

60%

Faster report render times

100%

Purview governance coverage

Account profile

Industry: Business Services / Professional Services

Organization scale: Global enterprise, $750M+ revenue

Tech stack:

Microsoft Fabric (OneLake, Warehouse, Lakehouse)
Azure Synapse Analytics (legacy)
SSIS / Azure Data Factory (legacy)
Power BI (DirectLake)
Microsoft Purview
PySpark / Spark Notebooks
Entra ID / RBAC / RLS

When five platforms become a liability

Growth through acquisition left each business unit on a different Azure analytics stack. A decade of project-by-project decisions produced a fragmented data estate spanning multiple jurisdictions, and no single team was able to govern, optimize, or trust it.

Microsoft's forced migration deadline

Microsoft retired Power BI Premium P-SKUs and feature-froze Synapse dedicated SQL pools. The organization had 90 days to commit to a migration path—with no internal Fabric expertise, no plan, and no visibility into dependencies across five Azure services.

Five disconnected services, 200+ SSIS packages

The organization ran Synapse SQL pools, ADF, on-premises SSIS, Power BI Premium with legacy SSRS, and ADLS Gen2—all without unified governance. Over 200 SSIS packages encoded years of business logic in procedural ETL that couldn't be lifted and shifted. Batch pipelines ran without observability, duplicated data produced conflicting KPIs, and cross-border data residency requirements went unmanaged.

Six-week analytics backlog, unreliable forecasting

Analytics requests were backlogged six-plus weeks. Finance couldn't reconcile cross-division revenue. Unpredictable Azure costs made quarterly forecasting unreliable. The C-suite had zero visibility into per-unit data infrastructure spend, and the data team spent more time maintaining pipelines than delivering insights.

From fragmentation to a governed lakehouse

Pythian started with the business, not the migration. Every workload, dependency, and piece of embedded logic was mapped before a single table moved. The goal was to build a platform that delivers governed self-service analytics and positions the organization for production AI—without disrupting operations.

Discovery phase

We cataloged 200+ SSIS packages by complexity, mapped Synapse stored procedure dependencies, audited ADF pipelines for Fabric feature-parity gaps, inventoried SSRS reports, and profiled Power BI models for DirectLake compatibility. We also mapped data residency requirements across 30+ countries to ensure jurisdictional compliance. The output was a migration roadmap sequenced by business value.

Strategic architecture

We designed a production-grade OneLake architecture with medallion layers and governance built in from day one:

OneLake Medallion layers

Bronze, Silver, and Gold layers with a workspace hierarchy mapped to business domains and regional data residency boundaries.

Purview governance

Classification, sensitivity labels, lineage tracking, and policy enforcement across the entire Fabric tenant.

F-SKU capacity model

Right-sized capacity with per-business-unit chargeback visibility, replacing the opaque Premium P-SKU cost structure.

Implementation roadmap

Phase 1: Assessment and architecture (weeks 1–6)

We conducted a full-estate audit of Synapse, SSIS, ADF, SSRS, and Power BI workloads, and scored every SSIS package for complexity. We mapped data residency requirements, designed the OneLake architecture with Purview governance and an F-SKU capacity model, and delivered a prioritized roadmap to the CIO and CTO offices.

Phase 2: Migration and modernization (weeks 7–18)

We migrated Synapse SQL pools to Fabric Warehouse with DDL refactoring and T-SQL remediation, and re-engineered SSIS packages into Fabric-native ELT. We converted ADF pipelines and transitioned Power BI Premium to F-SKU capacity with DirectLake optimization. Every workload went through dual-run validation to ensure zero disruption.

Phase 3: Analytics, AI readiness, and managed services (weeks 19–24)

We converted legacy SSRS reports to self-service Power BI dashboards and consolidated redundant semantic models. We implemented CI/CD pipelines with Git integration and enabled Fabric notebooks for the data science team. We also initiated a 24/7 managed services engagement covering capacity monitoring, Spark tuning, refresh health, and security alerts.

Measurable outcomes from a unified platform

Consolidation didn't just reduce complexity. It fundamentally changed how the organization uses data—shifting from reactive report requests to governed self-service analytics, with an AI-ready foundation that is already in use.

35% reduction in Azure Data Platform spend

Pythian consolidated five services into one capacity model and eliminated redundant compute and storage. Per-unit chargeback visibility gave the CFO accurate quarterly forecasts for the first time.

60% faster report render times

Report render times improved by 60 percent after migrating to DirectLake. Business users now get answers in seconds, not hours.

40% faster data pipeline execution

Spark tuning and V-Order optimization cut pipeline execution by 40 percent, replacing the rigid batch schedules and opaque failures that had consumed the data team's time.

100% Purview governance coverage

Classification, lineage, and policy enforcement across all data assets. Purview policy automation replaced manual compliance oversight across 30+ country operations.

First production ML model in 60 days

The analytics backlog dropped from six-plus weeks to same-day turnaround, and the data science team shipped its first production ML model within 60 days of go-live.

Microsoft Fabric consulting services

Ready to solve your data challenges?

Speak with a Pythian Fabric expert ->

On this page

Ready to unlock value from your data?

With Pythian, you can accomplish your data transformation goals and more.

Speak with Pythian consultants now →

Services