GigaCarbon | Route Proof

Illustrative workflow
The problem we are building for
Not Route Proof output
Route Proof today reads one historical source per run

Top-down view of a complete semi truck and trailer

Records

Consolidated operating record

Illustrative

Vehicle	Date	Source	Distance	Fuel	Event
UNIT-014	14 AUG	TELEMATICS	184.2 MI	—	TRIP
UNIT-014	14 AUG	FUEL CARD	—	62.4 L	FUEL
UNIT-027	14 AUG	MAINTENANCE	—	—	SERVICE
UNIT-031	15 AUG	TELEMATICS	96.8 MI	—	TRIP

Somebody checks this by hand

Telematics exportAB12 CDE14 AUG184.2 MI

Fuel card exportAB12CDE14 AUG62.4 L

Same vehicle?Somebody still has to decide.

Illustrative

Records support decisions

FinancialPay / challenge an invoice

OperationalRepair / investigate / schedule

Sustainability / capitalReport fuelReplace an assetJustify investment

GigaCarbon

The decision is only as defensible as the records beneath it.

Operational teams already make decisions from historical records. The harder question is whether those records actually contain enough evidence for the conclusion being made.

Today, Route Proof reads one historical source at a time and reports what that record can support, what it cannot, and what is missing.

The core idea

A record is not simply good or bad. It is sufficient or insufficient for a particular question.

Evidence sufficiency depends on the question.

What we've actually tested

A research ledger, including where the method fails.

We test the method against different kinds of operational records — including cases where it supports a conclusion, refuses one, changes a classification, or exposes a failure in our own method.

Test	Question	Outcome / status
Leeds	Can the published records independently support fuel-efficiency comparison?	0 / 1,482 independently recomputableAn earlier ~83% claim was retracted.
Mesa	Can the same method affirm a narrower conclusion when the evidence is actually present?	110,976 eligible records supportedFor the narrower aggregate source-reported-quantity question. Positive control
Cincinnati	Can evidence requirements materially change an operational classification?	14 / 134 active assets changed classificationUnder the evidence-bounded treatment. This does not imply paired-source reconciliation.
Synthetic GPS	What should happen when movement occurs during a recording gap?	95.3 km observed126.1 km endpoint displacement kept separate. Total route distance not established. Synthetic · self-scored engine test · not customer validation
Madagascar	Can the method identify consequential workbook problems and support decision-triage?	MixedThe method located consequential workbook problems but failed decision-triage.
WCC	Could the required evidence be retrieved sufficiently to run the intended benchmark?	Negative feasibility resultRequired evidence could not be retrieved sufficiently to run the intended benchmark.
Epistemic-restraint protocol	Registered protocol awaiting execution.	Registered / unrunNo result yet.

One of these tests changed the method.

Public data
Method demonstration

Leeds

One real file

1,482 records

Two questions

We originally reported that roughly 83% of the fleet could be compared on fuel efficiency. The analysis was reproducible. The conclusion was not supported.

What the source reports1,259 / 1,482Reported by source

What we can independently recompute0 / 1,482Not established

The published record did not contain the per-vehicle fuel quantity required to recompute fuel efficiency.

The original ~83% claim was retracted.The failed output was retained as a regression fixture.

What changed after Leeds

We now separate what a source reports from what we can independently recompute. If a conclusion requires evidence the record does not contain, the result is reported as not established rather than filled in by assumption.

The failed analysis was retained as a regression fixture so the same mistake is tested against going forward.

Product Evidence Method

Contactcontact@5ceco.com