Data Foundations
ManufacturingUse case, not a client storyBring every machine on the floor into one data store.
Your machines already record temperature, pressure, cycle time and stoppages. That data sits in each PLC, a local historian or a USB stick, and none of it lines up with your orders. We bring it into one store your team can query.
Every machine, one storeSchematic
Press line PLC
CNC controller
Furnace sensors
Energy meters
One data storeRaw signals, keptCleaned and timestampedJoined to the business
Joined to work orders, shifts, batches, maintenance logReady to query
4.6×Target return in 12 months
Week 1Fixed-fee audit
6 weeksTo the first live result
3Senior people on the work
TypeUse case: what we build, not a client story
ForPlant heads, operations and maintenance teams in mid-sized or small factories
EngagementWeek 1 audit, then build in about six weeks, fixed scope, fixed fee
Runs onAntvia lakehouse in your own cloud, on open formats
The problem
The data exists. Nobody can use it.
Ask why line 3 was slow last Tuesday and the answer takes a week. Someone walks to the machine, copies a log, opens it in a spreadsheet and tries to match it by hand to the shift and the order. By then the question has moved on.
What was broken
- Each machine keeps its own data, in its own format, on its own clock.
- Older machines record nothing at all unless someone adds a sensor.
- Readings carry a tag name, not the order, batch or shift they belong to.
- Logs are overwritten after a few days, so history is lost.
- Every analysis starts with someone copying files by hand.
You might be here if
- You cannot say which machine stopped most last month, and why.
- Your maintenance team works from memory and paper logs.
- A quality complaint means days of digging to find the batch conditions.
- You have bought sensors, but their data goes nowhere useful.
Why it mattered. Downtime, scrap and energy cost are decided on the floor. If the data from the floor cannot reach the people deciding, every improvement is a guess.
The approach
Start from the questions, not the machines.
We do not wire up every tag on day one. In Week 1 we walk the floor with your plant and maintenance leads, list the questions that cost you money, and pick the few machines that answer them. Then we connect those first, prove the answers, and add the rest line by line.
On the floor
PLCs and controllers
Sensors and meters
Orders, shifts, batches
woodfrog method
Pick the questionsWith your plant and maintenance leads
Connect and land rawA small gateway per line, readings kept as sent
Clean and joinOne clock, one unit, matched to the order
What changes
Full machine history
Readings tied to orders
Answers in minutes, not a week
Deliberately not doneReplacing your PLCs, or connecting every tag before anyone has asked a question
The solution
One store, every machine, joined to the business.
What we build
- Read-only connectors to your PLCs, controllers and meters, using the protocols they already speak, such as OPC UA, Modbus or MQTT.
- Add-on sensors for older machines that record nothing today, where a question needs them.
- A raw layer that keeps every reading as sent, so history is never lost and any analysis can be replayed.
- A cleaned layer with one clock and one unit per tag, with gaps, spikes and stuck sensors flagged.
- A business layer that ties each reading to the work order, shift, batch and maintenance event it belongs to.
- A first set of reports your team asked for in Week 1, such as stoppages by machine and reason, and conditions per batch.
It runs on Antvia, a bronze, silver and gold lakehouse plus BI, in your own cloud and on open formats. You own the data and the code. If you run Uloborus for manufacturing operations, orders and batches come straight from it.
How it would runBuild runs from Week 2 to Week 6. Operate is open-ended and drawn at one week only to fit. More lines are added after the first ones prove the answers.
Wk 1Wk 2Wk 3Wk 4Wk 5Wk 6Wk 7
Audit
Questions, machines
Connect
First lines, raw layer
Clean and join
Clock, units, orders
Reports
First answers live
Operate
Ongoing
Wk 1Wk 2Wk 3Wk 4Wk 5Wk 6Wk 7
Audit
Questions, machines
Connect
First lines, raw layer
Clean and join
Clock, units, orders
Reports
First answers live
Operate
Ongoing
The impact
What we would measure.
From a machine signal to an answerEach reading passes through three layers before anyone builds a report on it.
Schematic, no valuesBronzeRaw signals, as sentEvery tag, every reading, kept so you can replay it
SilverCleaned and alignedOne clock, one unit per tag, gaps and spikes flagged
GoldJoined to the businessReadings matched to the order, shift and batch they belong to
The raw layer is never thrown away, so a question nobody asked yet can still be answered later.
What we would measure on a real engagementWhat we would measure
Machines connectedShare of the floor sending data to one store
Time to answer a floor questionFrom the question to a trusted answer
Readings tied to an orderShare of readings matched to an order, shift and batch
Unplanned stoppagesBy machine and reason, against the Week 1 baseline
Return on the workTarget: at least 4.6× within 12 months, measured on your own numbers
Get your machines talking to one store.A short call, or record a message. A person replies within one business day.
