Make the result traceable

Match

The supplied net billing equals the confirmed expected amount for the selected agreement and period. This does not prove service delivery or legal compliance.

Potential discrepancy

The amounts differ. Show whether the billing is potentially below or above the reference, together with the records and calculation. It is not money recovered.

Accepted exception

A customer-confirmed waiver or other approved commercial decision changes what should be expected. Preserve its supporting record rather than treating it as an error.

Missing evidence

The expected amount or a required condition is unconfirmed. Leave the difference unassessed. An unknown amount must not become zero.

Each finding names the agreement clause it relies on and the invoice or credit record it was compared with. The same line shows the period and the review status, with the calculation beneath. The customer confirms the relevant terms before anything is compared.

The sample references are authored illustrations. No source PDFs were extracted or supplied by a customer. Its amounts exclude GST for a like-for-like comparison.

A small deterministic Python prototype was run against developer-authored synthetic inputs: structured, pre-confirmed schedules and allocated billing records.

These checks cover specified examples such as fixed fees, credits, repeated imports, split invoices and missing evidence. They are not an independent audit, an accuracy estimate, AI extraction testing, production reliability evidence or a customer outcome.

The fictional A$3,120 expected fee is supplied to the prototype. It does not extract a 4% clause from a contract.

The prototype does not read PDFs, establish a change's legal effective date, discover omitted agreements, resolve amendment precedence or verify service delivery. Safe live-data handling and automatic onboarding have not been established by these tests.

The website sample is a separate demonstration of three fictional records. It does not claim to implement every Python scenario or process customer files.

A real finding needs review by the customer or their authorised adviser. Missing evidence must be obtained before a supported conclusion can be reached.