A readout to a client's leadership team
Case study

A consumer goods manufacturer cut monthly consolidation from 120 hours to 30

Each acquisition brought its own ERP, so consolidating the month took weeks of analyst time and nobody saw inventory across sites.

Finance and reporting

75%less analyst time on monthly financial consolidation

The workflow before#

A fast-growing consumer goods manufacturer was buying companies, and each acquisition arrived with its own ERP: QuickBooks at one, Syspro at another, Sage at a third, on top of the SAP system the company was slowly moving to. Finance analysts closed each month by exporting from every system and reconciling in spreadsheets, about 120 hours of work. Operations couldn't see inventory across warehouses in one place, so buyers ordered on partial information. Every new acquisition made it worse, because its data had to be stitched in by hand before anyone could compare it with the rest. The SAP migration would fix this eventually, but it was years from done.

What we built#

We built the manufacturer one lakehouse on Databricks that reads from all five of its ERP systems and gives finance and operations a single set of numbers. It doesn't wait for the SAP migration and doesn't fight it. Each system keeps running, and the lakehouse does the reconciling that analysts used to do in spreadsheets. When the company buys another business, its ERP plugs into the same pattern.

Five ERPs, one set of tables

Data from SAP, QuickBooks, Syspro, Sage and a legacy system lands raw, then gets cleaned and mapped to a shared model of products, customers, sites and accounts. We mapped to the company's future SAP structure rather than any one legacy system, so the work done now carries straight into the migration.

Close-ready finance data

Validation rules check every load before it reaches the finance tables, so analysts start the close from reconciled figures instead of raw exports. Monthly consolidation went from about 120 hours to 30. The checks run on the data, not on the analyst's memory of which export had the wrong currency.

Inventory across every site

Power BI dashboards show stock across all warehouses on one screen, refreshed from the same tables finance uses. Buyers and planners see what's on hand everywhere before they order, and report users grew from about 50 to 250 as teams found they could answer their own questions.

Access rules that travel with the data

Unity Catalog holds the access controls and lineage centrally, so each team sees what it should and an auditor can trace any figure back to its source system. We set governance up as its own phase because an acquisitive company keeps adding users who shouldn't see everything.

How we delivered it

A team of solution architects, data engineers and analytics specialists worked with the client's IT and finance staff over six months. We started with a pilot on the largest warehouse site and primary ERP, then added the other systems one at a time with parallel quality checks, then governance, then tuning, automation and training.

Before and after#

MeasureBeforeAfter
Monthly financial consolidationAbout 120 hoursAbout 30 hours
ERP systems in one reporting viewNone; reconciled in spreadsheetsAll five
Active report usersAbout 50About 250
Inventory visibilityPer site, by requestAll warehouses on one dashboard

What's next#

The shared data model is already shaped like the target SAP structure, so the lakehouse becomes the validated source for the migration's data loads. It's also the fastest route for the next acquisition: a new ERP maps into existing tables instead of starting another round of spreadsheets.

ShareLinkedInEmail

Azure Data Factory orchestrates extraction from SAP, QuickBooks, Syspro, Sage and a legacy ERP into Azure Databricks. A medallion architecture on Delta Lake handles the flow: Bronze holds raw ingested data with ACID guarantees, Silver applies PySpark and SQL cleansing, standardization and cross-system mapping, and Gold holds business-ready finance and inventory datasets for Power BI.

Unity Catalog provides centralized access control, role-based permissions and lineage. Databricks Asset Bundles manage CI/CD and infrastructure as code, so new sources and environments deploy the same way. Validation rules run on each load before data is promoted to Gold.

Delivery ran in four phases over 24 weeks: foundation and pilot ERP (weeks 1 to 4), sequential integration of the remaining systems with parallel validation (weeks 5 to 12), Unity Catalog governance (weeks 13 to 16), and performance tuning, automation and training (weeks 17 to 24).

Next step

Let's build the case study we tell next.