What makes an extracted document field usable for finance?
Choose a fictional invoice with a total, a tax amount and one ambiguous line. Ask the reviewer to trace the fields to their source excerpts, correct the ambiguous value and compare the resulting CSV or JSON with the approved record.
Document extraction or automatic accounting entry?
Start with the output you need to use. The important test is whether a reviewer can trace each extracted value back to its source before exporting it.
Use these criteria to prepare your product evaluation.
Decision
A good fit when
Confirm first
Repeated documents
You need the same invoice, receipt or document fields in a usable file.
Test a clear sample and a difficult sample; compare missing values, dates and amounts.
Tables and exports
You need rows and columns in CSV or structured JSON.
Open the export in your target tool and check headings, decimal values and row alignment.
Review and retention
Someone will check AI output before it enters business records.
Agree the reviewer, export location and retention process; confirm the published file and page limits.
Review a supplier invoice before exporting its rows
A fictional invoice has six units at £15 and a £12 delivery charge. Its subtotal is £102 before tax.
Extract the fields and rows, check the quantity and unit price against the document, then review the export in a spreadsheet.
What to check The export preserves the separate delivery line and the item calculation. The reviewer can trace the £102 subtotal back to the source.
Build a useful evaluation
Five practical ways to evaluate VCMP Unfold
Start with the task that matters to your team. Each brief gives you an example to prepare, a workflow to review and a result to interpret. Use fictional or anonymised records in a demonstration.
Evaluation 1 of 5
Extract invoice fields with source checks
An invoice extraction is useful when the reviewer can trace important values back to the document.
Prepare
Use a fictional invoice with a supplier name, invoice number, date, tax and total. Include one deliberately ambiguous line.
Check
Compare the extracted fields with their source excerpts, correct the ambiguous value and check the final total before export.
Read the result
Extraction is not accounting approval or automatic posting. Retain the source document in your own records as required by your process.
Evaluation 2 of 5
Review tables before CSV export
A table can look plausible while a wrapped description or missing row changes its meaning.
Prepare
Prepare a small fictional table with a multi-line description, a blank cell and a numeric identifier with leading zeros.
Check
Review the extracted rows and columns, edit errors and open the CSV in the destination spreadsheet with suitable column types.
Read the result
Spreadsheet software may reinterpret dates and identifiers. Compare the exported values with the reviewed document before using them.
Evaluation 3 of 5
Define custom fields around a decision
A custom extraction should capture the information needed for a business decision rather than every phrase in the document.
Prepare
Choose a fictional contract and specify the counterparty, renewal date and notice period, including how missing values should be handled.
Check
Configure the requested fields and review each result against its source. Mark an absent value as missing instead of accepting a guess.
Read the result
AI output needs human review. An extracted date or clause summary does not establish the legal meaning of a contract.
Evaluation 4 of 5
Batch documents without losing exceptions
A batch saves handling time only if each document remains identifiable and errors stay visible.
Prepare
Prepare three fictional files: a clear PDF, a scanned receipt and an unsupported or oversized sample for the agreed test.
Check
Check upload acceptance, document identity and extracted results separately. Review failed or incomplete documents before treating the batch as finished.
Read the result
Check Unfold's current file-size and PDF-page limits. Batch completion does not mean every extracted value is correct.
Evaluation 5 of 5
Choose export and retention requirements
A useful extraction workflow ends with a reviewed result in the system where the team actually works.
Prepare
Identify whether the recipient needs CSV, JSON or Markdown and which fields it expects. Agree who will retain the reviewed export.
Check
Export a fictional result and inspect it in that destination. Check Unfold's current result-retention period before relying on history.
Read the result
Unfold does not automatically post results into accounting. Original files are processed in memory; provider processing and retention terms still need review.
Bring your most important workflow to the next step. Note the country, languages, devices and evidence your team needs.
Use sample records in a demonstration or trial. Pick the checks that matter to your team and record a pass, gap or follow-up alongside what you observed. Confirm which checks apply to your plan and configuration.
Prepare an extraction
Choose the invoice page that contains the source of a disputed amount.
Use a multipage sample table whose column header repeats and inspect the extracted rows.
Compare field extraction for a digital PDF and a scanned test document.
Check how an empty source field is represented in the resulting record.
Inspect a line that wraps across two printed rows before treating it as two items.
Choose which document fields the downstream sheet actually needs.
Use synthetic documents to test whether similar supplier names stay distinct.
Verify source and result
Compare a date field with the format in the original document.
Check that a reference containing leading zeros survives extraction.
Verify that a negative adjustment remains negative in the result.
Compare two currencies in one sample document before summing values.
Reconcile one quantity and unit cost against the extracted line subtotal.
Check the extracted source page for a clause that spans a page break.
Record a correction and retain the original source for independent review.
Export and handover
Open a sample export and check that decimal separators import as expected.
Check whether the exported source reference lets another person find the right page.
Confirm that multiline notes remain in one spreadsheet row.
Check how duplicate document references appear when combining exports.
Validate a totals calculation after importing a corrected line.
Compare export column names with the receiving system’s required columns.
Identify who checks confidential fields before an export leaves the team.
Small businesses handling invoices, receipts, contracts and other paperwork.
Invoice and receipt extraction
Document tables to CSV
Source checks before using AI data
What does it cover?
PDF, Word, image and text uploads
Preset and custom fields
Source evidence and editable review
Batch uploads, history and exports
How it fits the family
Unfold focuses on getting useful, reviewable data out of documents. It has its own workspace and does not automatically post results to accounting or other VCMP products.
What to confirm before choosing
Check the 3 MB and ten-PDF-page limits. AI results need human review. Results remain available for 30 days.
Buyer questions, answered
Find a question (15)
Can I try Unfold without paying?
The verified-email trial includes 25 documents for seven days and needs no card. Paid plans are £19 for 100 documents or £49 for 400 documents per calendar month.
VCMP processes original files in memory and retains extracted results for 30 days. Document contents are sent to OpenAI for processing; its applicable provider retention can still apply.
How do I choose between document extraction and a complete invoice automation system?
Choose extraction when the main task is to turn document fields or tables into reviewed CSV or JSON data. Unfold includes source excerpts and editable results for that step. If you also need invoice approval, supplier matching or posting to an accounting system, include those requirements in your comparison and confirm the connections separately; extraction alone does not complete a finance workflow.
How should I assess an AI extraction tool's accuracy on my own invoices?
Use representative fictional or appropriately authorised invoices, including an unclear scan, a missing value and a table with several rows. Compare extracted names, dates, currency, tax and totals with the source, then review the exported columns. Unfold provides source evidence and editable review. Treat missing or ambiguous results as review tasks rather than assuming a model's output is correct.
What should I check before using extracted document tables in a spreadsheet?
Check the column names, row order, decimal and date formats, missing cells and source references before using the export. Unfold offers CSV and JSON results, with a 3 MB file limit and a ten-page PDF limit. Download the results you need within the 30-day retention period, and review totals against the source before the spreadsheet informs a business decision.
What sample documents should we use in an Unfold evaluation?
Choose different layouts, one multi-page example and one document with an unclear value. Compare the required fields and tables with their source excerpts.
The product guide lists a 3 MB file limit and a ten-page PDF limit. Confirm your representative documents fit before planning a larger extraction workflow.
How can we check whether extracted tables will fit a spreadsheet?
Export a reviewed sample as CSV and open it in the receiving spreadsheet. Compare column names, row boundaries, dates and numeric values with the source.
Does an Unfold JSON export create an accounting integration?
No. JSON is an export format. Agree the receiving system, field mapping and reviewed import or integration separately before expecting records to appear elsewhere.
How should we test a document with repeated headers or totals?
Use a representative multi-page sample and compare each row with its source. Check that repeated headers and summary totals are not mistaken for additional transactions.
Results remain available for 30 days under the published scope. Agree a review and export routine that preserves the files your business needs before access expires.
Who should sign off extracted business data before reuse?
Nominate a reviewer who understands the source and receiving workflow. Have that person resolve ambiguous values and approve the reviewed export before downstream use.
How should we compare document extraction accuracy fairly?
Use the same documents and required fields across trials. Count correct, incorrect and missing values separately, including the human review time needed to produce an accepted export.
Can Unfold be evaluated without uploading confidential customer files?
Yes. Use synthetic or appropriately redacted samples with representative layouts. Agree document-processing requirements before introducing confidential material into an operational workflow.
This guide is maintained by StartMadeSimple Ltd using its public product information. Read the official source for VCMP Unfold. Confirm current prices, access and configuration directly before purchasing.
Review a demonstration document, compare extracted values with source evidence and inspect the export choices. Use the demonstration records to evaluate the workflow.
Field extraction accuracy alone hides the cost of checking ambiguous documents. Compare source traceability and correction effort before accounting automation.
Published scope, with a practical decision test
VCMP Unfold
Nanonets
Unfold extracts reviewable fields and tables with source excerpts and CSV or JSON exports.
Nanonets publishes AI document and business-process automation.
Test the difference on your own workflow
Use a synthetic invoice with a credit line and an ambiguous date; extract its totals and check them against the source.
What to record: Preserve the original meaning of the credit, correct uncertain fields and inspect the exported values. Confirm limits, retention and downstream integrations.
This is a suggested evaluation, not a measured customer result or a claim of complete feature parity.
What makes the best Nanonets alternative for your team?
The best fit is the option that completes your essential workflow with acceptable cost, access and support. Use the test above with the same inputs, then compare current written terms. A feature omitted here is not evidence that either supplier lacks it.
Vendor-authored by StartMadeSimple Ltd, the company behind VCMP. Public sources reviewed . Competitor products have not been exhaustively tested.
We use optional analytics cookies to understand which pages people visit and how they discover VCMP. We do not use these analytics cookies unless you accept them.