Evidence walkthrough
Two systems can describe the same event differently.
A compact synthetic example of how GigaCarbon preserves disagreement, structures the evidence, and leaves the judgement with the operator.
≠
Both records exist. Somebody still has to establish what actually happened.
Structure the evidence before the judgement.
Original assertionsPortal → Vehicle X
Field → Vehicle Y
Field → Vehicle Y
ConflictVehicle attribution disagrees
ContextVehicle X broke down.
Backup Vehicle Y was used.
Backup Vehicle Y was used.
StatusHuman judgement required.
Original records remain preserved.
Operator accepts Vehicle Y.
Accepted stateVehicle Y
Provenance trail remains visible
Portal originally recorded → Vehicle X
Field / operational record → Vehicle Y
Context: Vehicle X broke down and Vehicle Y was the backup.
How did this clean value become clean?
GigaCarbon keeps the path reconstructable.
GigaCarbon prepares the evidence.
The operator still makes the judgement.
Where we're testing it
We’re testing the same approach in several operational workflows.
Fuel reconciliation
Portal and field records disagree before an accepted operating record can be used downstream.
Repair / invoice review
Several records must be checked before money is approved.
Fleet analysis
A consultant needs to establish what client records actually support before applying professional judgement.
What we are testing
The question is not whether messy records exist. They do. The question is whether GigaCarbon can reduce the work required to turn those records into traceable, usable evidence without hiding uncertainty or taking judgement away from the expert.
Does it reduce manual reconstruction?
Does it preserve the reasoning behind accepted corrections?
Would the practitioner use it on the next case?
Public testing
Research that got us here
Earlier Route Proof work tested whether fleet records could support specific conclusions. Positive, negative and corrected results were retained. That work led to the current hypothesis: structure the evidence and preserve uncertainty before the expert makes the call.
Leeds
Correction retained. 0 of 1,482 records were independently recomputable for fuel efficiency. An earlier approximately 83% claim was retracted.
Mesa
Positive control. 110,976 eligible records supported a narrower aggregate source-reported-quantity question.
Cincinnati
Classification changed. 14 of 134 active assets changed classification under an evidence-bounded treatment.
Synthetic GPS
Unsupported total refused. Observed distance and endpoint displacement remained separate; total route distance was not established.
Madagascar
Mixed result. Consequential workbook problems were identified, but decision-triage failed.
WCC
Negative feasibility result. Required evidence could not be retrieved sufficiently to run the intended benchmark.
First review
One case. Narrow scope.
A first review can be structured so records remain in your environment, or use a redacted/minimised case or small approved extract. We do not ask anyone to breach an NDA or internal data policy.
How first reviews can work →Work with us
Have a workflow where the records don’t reconcile cleanly?
We’re looking for real cases where someone already has to resolve conflicting or incomplete operational evidence before a decision can proceed.
Show us one case