DataVance · Business advancement through data
Accountable data-automation APIs for Australian business
Every DataVance product turns your business data into a measured, maintained outcome — accuracy published openly, not just claimed. Our first, DVAP, automates accounts payable end to end.
Measured accuracy
Proof from our first product, DVAP — measured on the public SROIE benchmark, not claimed.
- Invoice-date accuracy
- 100%
- Total-amount accuracy
- 99%
- Straight-through rate
- 87%
- Average confidence
- 0.97
One library, one standard
Every DataVance API hangs off the same account and carries the same accountability — accuracy measured and published, an SLA and changelog behind it, data resident in Australia.
One account
DVAPLive
DVAP FSLive
DVAP DCLive
DVAP PMLive
DVEXLive
DVEILive
DVTXLive
Every product ships with
Measured accuracy, publishedAn SLA and a changelogData resident in AustraliaWhat every DataVance product shares
One standard runs through the whole library — buy the outcome, not another thing to build and maintain.
Accountability, not just code
Where being wrong costs money — tax, compliance, payments — every DataVance product ships with an SLA, a changelog, and someone whose job is keeping it correct as the rules change.
Measured correctness
Accuracy isn't claimed, it's measured against public and real-document benchmarks — and every per-field number is published openly. The evals are the pitch.
In your workflow, not a repo
Multi-tenant keys, quotas, exports, and signed webhooks mean you integrate in an hour and stay in the tools you already use — no software to host or debug.
The API library
Seven APIs live today, one account across all of them — invoices, receipts, e-invoice conformance, and Australian business identifiers, each sold and priced on its own.
- LiveDVAP
Invoice Intelligence & AP Automation
Extraction with per-field confidence, fraud and duplicate screening, PO matching, approvals, and ERP export.
- LiveDVAP FS
DVAP Fraud Screen
Score any supplier invoice for fraud risk — a 0–100 score with the reasons named. A focused, single-purpose slice of DVAP.
- LiveDVAP DC
DVAP Duplicate Check
Catch a duplicate invoice before it's paid twice — exact-file and same-details matching over every document you've checked. A focused, single-purpose slice of DVAP.
- LiveDVAP PM
DVAP PO Match
Match an invoice to its purchase order in one stateless call — 2-way and 3-way, every discrepancy named. A focused, single-purpose slice of DVAP.
- LiveDVEX
Expense & Receipt Extraction
Receipt processing — merchant, ABN, GST, and expense category — on the same measured-accuracy engine, with a confidence score on every field.
- LiveDVEI
e-Invoicing (Peppol / PINT A-NZ)
Validate a UBL e-invoice against the PINT A-NZ ruleset and get a conformance report that names every error, its rule, and where it is.
- LiveDVTX
ABN & ACN Validation
Check an Australian business identifier in one call — checksum, register status, GST registration, and how well the name matches.
Five more are planned — see what we build next.
What each one takes, returns, and keeps
The fastest way to pick: match the document you have to the answer you need. The last column is what survives the request.
| API | Takes | Returns | Keeps |
|---|---|---|---|
| DVAP | Supplier invoices | Fields, validation, fraud, PO match | Stored record |
| DVAP Fraud Screen | Supplier invoices | Risk score + named reasons | Nothing stored |
| DVAP Duplicate Check | Supplier invoices | Duplicate verdict + rule | Fingerprint only |
| DVAP PO Match | Invoice + purchase order | Line-by-line exceptions | Nothing stored |
| DVEX | Purchase receipts | Fields + expense category | Nothing stored |
| DVEI | UBL e-invoices (XML) | PINT A-NZ conformance report | Nothing stored |
| DVTX | ABN / ACN identifiers | Checksum + register status | Nothing stored |
What we build next
In the order we intend to build them. These are marked planned rather than in development because none has been started — you will see the badge change when that stops being true. If one of them would solve a problem you have now, telling us moves it up the list.
- DVRPPlanned
Resume Parsing
A CV in, structured JSON out — contact details, work history, skills, education, and qualifications, each with its own confidence score.
Document AI / HR
- DVTEPlanned
Document Text Extraction
Any document to clean, ordered text — the primitive under everything else we build, sold on its own for when text is all you need.
Document AI
- DVNEPlanned
Named Entity Extraction
Text to named entities — people, organisations, ABNs, amounts, dates — with the span each one came from, so every value is traceable.
AI / NLP
- DVDEPlanned
Document Entities
File to entities in one call: text extraction and entity recognition chained, for the outcome most teams actually want.
Document AI
- DVDCPlanned
Data Cleaning & Transformation
Messy records to consistent ones — addresses, names, dates, currencies — normalised with the rule that fired recorded against each change.
AI / Data
Tell us which one you need — a person reads every one of these.
The numbers, published
Measured by the benchmark run and versioned with the code — the same payload the API serves at /metrics/accuracy. Scores are never averaged across datasets, because a synthetic set and a set of photographed receipts do not measure the same thing.
Field accuracy on scanned receipts
SROIE, the public scanned-receipt benchmark — real photographed receipts, not our own documents. n = 100
Bars run 0–100%. The pale track is the remainder to 100%.
- Straight-through rate
- 87%
- Documents needing no human correction.
Fraud and duplicate detection
Synthetic golden set — 1,000 generated invoices with known planted fraud, so every catch and every miss is verifiable. n = 1,000
Bars run 0–100%. The pale track is the remainder to 100%.
- Fraud F1
- 0.96
- Precision and recall combined.