A production operations system
fresh-produce processing · West Africa · pilot, in progress
Digitising the operating layer of a fruit processing plant.
How it ran before
Fruit arrived and a supervisor wrote the supplier, variety and batch number onto a sheet clipped to a board. That sheet followed the batch through peeling and cutting, gathering corrections in three hands. Output was weighed and recorded on another sheet, in another book.
Reconciling what went in against what came out was an end-of-week exercise in arithmetic and memory. By the time anyone could see that a line had underperformed, the shift, the batch and the chance to intervene were gone.
How it runs now
Intake is captured once, at the point it happens, on a device that works whether or not the plant has signal. The batch carries its own identity from that moment: every allocation, every station, every weight recorded against it by the person doing the work. That makes a traceable line from supplier to finished pack, readable in either direction.
Reconciliation is a property of the system rather than a task. Yield, throughput and waste resolve as the shift runs, in front of the people who can still act on them. Management reporting is a read of the same record, not a re-entry of it.
What the system holds, stage by stage
01
Intake
Supplier, variety, batch number and condition captured once, at the gate, on a device that works whether or not the plant has signal.
02
Allocation
The batch carries its own identity into every split and transfer, so a pallet three stations later still knows where it came from.
03
Stations
Peeling and cutting recorded against the batch by the person doing the work, rather than onto a sheet that gathers corrections in three hands.
04
Weighing
Output recorded at the scale, at the moment it is weighed, into the same record that holds the intake it came from.
05
Reconciliation
What went in against what came out, resolving as the shift runs. A property of the system, not an end-of-week task.
Detail
The full engineering write-up.
How the data model, batch traceability, staff weighting and migration path were actually designed, and what the pilot does not yet claim.
- 2017
- Operating since
- 21projects
- Across industries
- 6years
- Longest system in use
The traceability chain the system holds
01
Physical operation
A delivery arrives, a line runs, a shipment sails, a vehicle is on the road. Something happens in the real world, whether or not anything records it.
02
Events
The moments in that work worth recording. An arrival, a transfer, a reading, a loss. Record too much and it becomes noise. Record too little and nobody can trust it.
03
Structured data
Each event is written down once, at the moment it happens, and linked to the thing it belongs to: a batch, an order, an asset, a policy. This is where traceability is created or lost.
04
System
The software or model that the work runs on. Not a second tool that someone types into later. If the system is not part of the work, the record will not match the work.
05
Decision
Yield is short. Stock will not last the month. This risk is priced too low. All of it comes from the same record that captured the work, early enough that somebody can still act on it.
06
Action
Somebody acts. They restart a line, change an order, adjust a price, or set a new limit. Every step before this one is only there to make this step happen.
Build with Vaarde
Complex operations deserve better software.
Tell us how the operation works today. We will come back with what we think the system should hold, and what it should be telling you by the end of each shift.