Skip to content
Chayan Aggarwal Résumé
All case files
Case 05AcademicDesign proposalITKM-505 / TCMG-526 group project, University of Bridgeport, February 2026

One source of truth for SAP data

An enterprise data design that carries SAP data through Bronze, Silver and Gold layers to executive dashboards.

The proposed data flow, from SAP source systems to executive dashboards.

The chart

  1. 01 / Symptom

    A manufacturer's SAP systems ran operations well, but leadership couldn't get consistent answers to questions that cross functions.

  2. 02 / Tests

    Mapped how SAP data spreads across modules and how each new report created another extract, then compared Azure services for each job.

  3. 03 / Diagnosis

    Reports were built on separate, inconsistent extracts with no shared, governed layer underneath.

  4. 04 / Prescription

    A medallion design on Azure: raw data kept in Bronze, cleaned in Silver with Databricks, modeled in Gold for Power BI, with governance built in.

  5. 05 / Outcome

    A design paper and presentation. It wasn't built or costed, so no performance results are claimed.

How the tests ran

  1. Bronze: scheduled Azure Data Factory pipelines land an unmodified copy of SAP data in ADLS Gen2, so any later step can be audited or re-run.
  2. Silver: Azure Databricks and Delta Lake cleanse, validate, deduplicate and join tables, and apply business rules.
  3. Gold: Azure Synapse Analytics models the data as a star schema, ready for Power BI dashboards.
  4. Governance is part of the design from the start: Microsoft Purview for lineage, Entra ID for access, Key Vault for secrets and Azure Monitor for observability.

Tools

  • SAP
  • Azure Data Factory
  • ADLS Gen2
  • Azure Databricks
  • Azure Synapse Analytics
  • Power BI

Findings

  • 3

    layers, each with one job

  • Ingestion, transformation and reporting can each change without breaking the others.

Limits

  • This is a design. It wasn't built, load-tested or costed against real data volumes.
  • The company and its data are hypothetical, so service choices are reasoned from documentation rather than measured.

Links

Next case / 06Forecasts that survive a shock