Work – Martin Grosse | Engineered Analytics

60+ daily users

5 business functions

One number volume plan to plan on

Challenge

A European consumer goods business was running an IBP* process across five functions – with five different views of the plan. Multiple ERPs, disconnected systems, and spreadsheet-driven reconciliation meant that when the business came together to review demand, supply, finance, and operations, the numbers depended on where you looked. The process was there. The discipline was there. The data wasn't.

Approach

I designed and built a unified planning data model on Microsoft Fabric – integrating the disparate source systems, transforming operational data into a single volume plan, and building the semantic model that powers IBP reviews at every level of the organisation. Full stack: lakehouse, pipelines, dimensional model, Power BI reporting layer.

Outcome

Sixty-plus daily users across five business functions now work from the same number. The IBP cycle – from operational reviews through to the Management Business Review – runs on one version of the plan. The dialogue has moved from reconciling the data to making decisions with it.

* Integrated Business Planning (IBP) is a business management framework – developed and licensed by Oliver Wight – that aligns every function of a business around a single operating plan, reviewed monthly at multiple levels of the organisation.

↑ Back to top

200+ users served

4 ERP source systems

Metadata-driven pipeline orchestration

Challenge

A decision to migrate from on-premise SQL servers to Snowflake left more than two hundred people in a difficult position. Supply chain, planning, marketing and operations teams had built reporting workflows on direct SQL connections – Power BI dashboards, Excel workbooks, custom queries. Snowflake offers no native SQL connector for Power BI or Excel. With IT planning to retire the on-premise servers, the user community would have been stranded: data moved, but access gone.

Approach

I designed and built a metadata-driven ingestion layer on Microsoft Fabric Lakehouse – pulling data from Snowflake across four ERP systems (three SAP instances, one JDA) and over two hundred source tables. Full and incremental loads coordinated centrally by a metadata control framework rather than individual pipeline scripts; sales order data refreshed hourly. Data lands in a three-layer architecture – Raw, Labelled, Modelled* – and is surfaced through Fabric SQL endpoints, giving existing Power BI reports and Excel workbooks a straightforward connection swap with no SQL rewrite required.

Outcome

The transition happened without breaking what already worked. Legacy reports switched from retiring on-premise connections to Lakehouse SQL endpoints – with a noticeable improvement in query performance. Supply chain planning users began moving from personal Excel/SQL workbooks to Excel connected to shared, governed data models. The platform became the data foundation for a broader analytics programme at the same client – including the IBP planning solution described above.

* Raw, Labelled, and Modelled are the three data layers used on this engagement – equivalent to the 'medallion architecture' (Bronze/Silver/Gold) pattern in modern data engineering. Raw holds ingested source data as-landed; Labelled applies business keys and definitions; Modelled serves as the analytical foundation for reporting and planning tools.

↑ Back to top

Impact.

One model. Five business functions. 60+ daily users.

Challenge

Approach

Outcome

Four ERPs. Two hundred users. Zero disruption.

Challenge

Approach

Outcome

Got a data challenge? Let's talk about it.