Building a Microsoft Fabric integration pipeline for CRM data

Context and objectives

The client needed to secure and mature the integration of its CRM data (HubSpot) into its Microsoft Fabric platform.

The data path itself added complexity: source data originated in the client's ERP system, moved through an SSIS integration layer and a SQL Server data warehouse, before reaching Fabric and finally HubSpot. An initial raw data layer was already in place, built during an earlier implementation phase, but the curated layers needed to make that data reliably usable were still missing.

Several challenges stood in the way:

• Identifying which CRM objects and fields were actually required, and how they should be mapped

• Reconstructing clarity on an existing but incomplete implementation, with limited documentation to build on

• Structuring a medallion architecture (raw, refined, and curated layers) on top of a fairly complex, multi-system data path

• Ensuring the pipeline stayed maintainable by a team without deep specialist knowledge of the underlying integration technology

Approach

1. Takeover and scope alignment

The project began by reviewing the existing Fabric implementation to understand what had already been built, identify gaps, and clarify a realistic scope going forward. This was followed by a dedicated scoping phase to validate, together with the client, which CRM objects and fields mattered and what integration rules should apply.

This work was captured in a shared reference document, giving both teams a common, structured basis for the technical build that followed.

2. Building the medallion architecture

With the scope validated, Agilytic built out the missing layers of a medallion architecture, taking the data from its raw form through to curated, business-ready datasets:

• Configured and stabilized the Fabric Dataflows and underlying pipeline steps needed to reliably move data from source to platform

• Designed and implemented the intermediate and curated layers on top of the existing raw layer, structuring the data around actual business consumption needs

• Resolved mapping and data-quality issues across the source systems involved

This gave the client a data foundation that was both functional and structured for long-term maintainability and governance, which was essential given the number of systems the data passed through before reaching HubSpot.

3. Stabilization and handover

A support phase followed, focused on troubleshooting, refining refresh behavior, and improving overall data quality, ensuring the pipeline could be trusted as a stable foundation rather than a one-off build.

The engagement closed with a structured handover, documenting what was delivered and identifying the next steps needed to mature the data product further.

Results

The project delivered a working Microsoft Fabric integration for CRM data, restructured around a proper medallion architecture:

• Took an incomplete platform from a raw data layer to curated, business-ready layers

• Implemented and stabilized the pipeline steps needed to move data reliably across a multi-system path, from the source ERP through to the final platform

• Delivered a validated data model and mapping document, giving the client clarity on scope and rules going forward

• Prepared concrete next-step recommendations to mature the solution further, including layer optimization and reporting readiness

The client valued the pragmatic approach and the structured alignment on scope and mapping. The engagement also surfaced further opportunities, including platform optimization and reporting enablement work.