A metadata-driven ingestion platform on Microsoft Fabric that lands data from fourteen of Feltex’s systems in one governed Bronze layer, ready for the analytics that come next.
Power Business Decisions with Cloud Scale Analytics
Microsoft
Delivery partner
Practice
The results
Feltex is one of South Africa’s leading automotive component manufacturers, a KAP business supplying acoustic, trim and foam components to the country’s vehicle makers and the aftermarket. Its operational data lived in many places: ERP and finance platforms, manufacturing execution and production monitoring, and HR and safety records kept in SharePoint, each reached a different way. Feltex wanted one controlled place for all of it to land, loaded the same way every time, with clear sight of what ran, what failed and what changed. It also needed the access control and South African data residency an enterprise platform has to have before anyone builds reports on it.
Every source follows the same path, and the register, not the code, decides what each run loads.
Sources
ERP, finance, manufacturing, production monitoring, HR and safety, over SQL, ODBC, REST, OData and shared folders.
Fabric pipelines
A thin parent pipeline per source holds only its identity and connection.
Ingestion framework
Four generic engines read what to load from one control register, so a new table is a row of metadata, not new code.
Driven by
OneLake
Source-aligned Delta tables, every record stamped with its source, pipeline and run.
SQL endpoints
Each lakehouse serves Silver, Gold, semantic models and reporting.
We designed and built the Bronze layer of Feltex’s data platform on Microsoft Fabric: the raw, source-aligned foundation every later layer reads from. Rather than write a pipeline per table, we built four generic, parameterised ingestion engines, plus a PySpark notebook for REST API data, and drive them all from a central control register. Each source has a thin parent pipeline that only knows how to connect. What to load, and how, lives in the register as data. Loads are incremental, keyed on watermarks that move forward only when a table lands successfully, so a failed load keeps its recovery point, and the largest systems run in sequenced load waves. Every record carries its source, pipeline, run and landing time, and every run is logged per stream and per table, with row counts for reconciliation. Access runs through Entra ID groups, scheduled work runs under a workspace identity, secrets live in Azure Key Vault, and settings for development and production sit in variable libraries so changes promote cleanly.
The controls an enterprise needs before anyone reports on its data, in place from the first run.
The Fabric environment runs in the Azure South Africa North region.
Admin, developer and viewer rights come from Entra ID groups that Feltex approves and reviews.
Scheduled pipelines and notebooks run under a workspace identity, not a person’s account.
API credentials sit in Azure Key Vault and are referenced through configuration, never kept in code.
Every run and every table logs its status, duration, row counts, errors and watermarks.
Each row carries its source, pipeline, run and landing time.
Counts recorded on every load support source-to-Bronze checks.
Watermarks move only on success, and failed or stale runs are retried and reconciled.
Feltex formally accepted the platform on 15 September 2026. It now has a single, controlled entry point for enterprise data: fourteen sources and 830 Delta tables landing in dedicated Bronze lakehouses in OneLake, each with its own SQL analytics endpoint. Onboarding a new table is mostly a matter of adding a row to the register rather than writing extraction code, and every load can be traced to its run, checked by row count and recovered from its last good watermark. We documented the platform and handed it over to Feltex’s own team, and it is now the governed base for the Silver and Gold layers, semantic models and reporting that Feltex builds next.
Download a one-page brochure of this story to share with your team.
Similar work in Industry & Infrastructure
Ivanhoe Mines
An end-to-end Microsoft Fabric and Power BI programme that moves Ivanhoe Mines’ reporting off spreadsheets and onto one governed platform, from geology and procurement to risk, finance and HR.
Power Business Decisions with Cloud Scale Analytics
Deep-shaft Mining Company
We mapped, optimised and governed a deep-shaft mining company’s HR processes in Nintex Process Manager, turning fragmented manual workflows into a navigable, automation-ready set of processes.
Practice
Nintex Process Manager
A modern .NET solution that prints product labels from automated quality checks on MAHLE’s South African lines, with central configuration and full traceability.
Practice
Application modernisation
Get in touch
Send us the gist of what you are trying to do. The right person from First Digital will be in touch within a working day.
