Skip to content
woodfrog
Data Foundations
Manufacturing and process industriesUse case on Databricks, no client result claimed

Put ERP and plant data in one lakehouse.

Production lives in machines and the MES, orders and costs in the ERP, quality in spreadsheets. We land them on Databricks so one set of numbers answers questions from the floor to finance.

On DatabricksSchematic
ERP, MES, PLC and quality sheets
Databricks lakehouse, joined and tested
OEE, yield and cost on the same numbers
RegisteredDatabricks partner, in Pune
CertifiedDatabricks practitioners on the team
6 weeksMedian to a first production result
Week 1Fixed-fee audit
TypeUse case: what we build on Databricks, not a client story
ForPlants running an ERP alongside an MES, PLCs and spreadsheets
EngagementWeek 1 audit, then build, fixed scope, fixed fee
Runs onDatabricks on your cloud, with change data capture from the ERP and streaming from machines
Before and after

From the old way to the new.

Before
  • Production, orders and cost live in separate systems
  • Reports stitched by hand each week
  • A machine number and a finance number that never agree
  • Machine data thrown away after a few days
After
  • One lakehouse holding orders, batches, machines and costs
  • Reports that refresh on their own
  • One definition of each number, from floor to finance
  • Machine history kept and ready to analyse
The problem

The plant knows what happened. Finance cannot see it.

Each system is right about its own corner. The PLC knows when a machine stopped, the MES knows the order, the ERP knows the cost. Joining them is manual work done once a month, so decisions rest on figures that are weeks old.

A yellow robot arm on a production line
Every machine and every order, in one place
What was broken
  • Machine, order and cost data never meet.
  • OEE is calculated one way on the floor and another in finance.
  • Quality results sit in spreadsheets.
  • Raw machine data is deleted to save space.
  • Every new question means a new manual extract.
You might be here if
  • Your weekly plant review starts with stitching exports.
  • Nobody can say what a stoppage cost in rupees.
  • You cannot compare lines or shifts fairly.
  • Your ERP is hard to change, so reports live outside it.

Why it mattered. Every stoppage and every scrapped unit has a cost, and you cannot reduce what you cannot see against the same numbers as finance.

The approach

Join on the order and the batch, then test the layers.

We trace one order from the ERP through the MES to the machines that made it, and note where the records join and disagree. Then we land each source on Databricks in layers: raw, cleaned, and business-ready, with tests at each step, so a number means the same thing everywhere.

Where it is today
ERP orders and costs
MES batches
PLC and quality data
woodfrog method
Trace one order end to endWhere records join and differ
Land each source in layersRaw, cleaned, business-ready
Test at every layerSo numbers stay trusted
What changes
One set of records
OEE and cost on one basis
Machine history kept
One lakehouse for plant and finance
Deliberately not doneReplacing the ERP or the MES. They keep running as they are, read only to start.
Read only to start: nothing writes back to the line until you approve it.schematic
The solution

One governed model from machine to ledger.

What we build
  • Connectors that read the ERP with change data capture, without slowing it down.
  • Streaming or batch ingestion of PLC and MES data, kept at full detail.
  • A layered model joining orders, batches, machines and costs on one definition of each number.
  • Data tests at each layer that stop bad data before it reaches a report.
  • OEE, yield and cost-of-loss tables ready for dashboards and analysis.
  • Governance with Unity Catalog, so each team sees what it should.
It runs on Databricks in your own cloud on open formats. Your ERP and plant systems stay exactly as they are.
How it would runWeek 1 ends with the data profiled and the scope fixed. Operate is open-ended and drawn at one week only to fit.
The impact

What we would measure.

What we would measureNo client values yet
Time to produce the weekly plant reviewBefore and after
Share of reports built by handCounted
Gap between floor OEE and finance OEEClosed to one number
Days of machine history keptBefore and after
Cost of loss per stoppageVisible by line and shift
See the plant and the ledger on one screen.A short call, or record a message. A person replies within one business day.
Contact usCertified Services Partner, Claude Partner Network