Fuel log auditor
A small system that matches fuel transactions to vehicles, works out which numbers cannot be right, and raises a query rather than an allegation. The interesting findings are almost never theft: they are data errors, a vehicle whose economy is quietly degrading, and one van that nobody has noticed is doing twice the mileage of the rest. Seven posts on the same system, one diagram at a time, with a cost breakdown and an engineering reference at the end.
- 01
A fuel log auditor on AWS for a few dollars a month
The whole system on one page — match, check, query — and why the framing of the output determines whether it is useful or corrosive.
- 02
How a fill gets matched to a vehicle
Why card-to-vehicle mapping drifts, how odometer readings get mistyped, the signature of a swapped pair, and what to do when the vehicle genuinely cannot be determined.
- 03
Why fuel economy only means something against itself
The factors that dominate fuel economy, why cross-vehicle comparison is nearly meaningless, how a per-vehicle baseline is built, and how seasonality is handled.
- 04
How a query gets raised without an accusation
Why the driver is asked first, the wording that gets a useful reply, why explanations must be recorded, and how the genuinely serious cases are handled differently.
- 05
What the fuel data says about the vehicles
How gradual economy decline predicts maintenance, the utilisation finding, how cost per mile should be computed, and what to report.
- 06
What the fuel log auditor costs
About $2 a month. One Bedrock read per transaction is the only line that grows; the queue, the table, the mail and the storage are rounding errors. Plus the three ways the bill could surprise you.
- 07
Engineering reference: the fuel log auditor architecture
Same system, drawn purely for engineers. Service names, region, Lambda inventory, IAM scopes, the schemas and the exact model id.
Frequently asked questions
- What is a fuel log auditor?
- A small serverless system that matches fuel card transactions to vehicles and odometer readings, identifies records that cannot be right, and raises queries with the driver before anything else.
- Is this a fraud detection system?
- No, and framing it that way makes it worse. Most discrepancies are data errors, and a system that treats every one as suspicion damages a workforce to recover very little.
- Why compare a vehicle only against itself?
- Because fuel economy varies enormously with load, route, weather and driver. A van against the fleet average tells you almost nothing; against its own history it tells you a lot.
- What is the most common finding?
- A fill recorded against the wrong vehicle, usually two vans with adjacent numbers. It looks alarming on both records and is a data problem.
- What does it cost to run?
- A couple of dollars a month. See part six.