Skip to content
SOIDEM Data Technologies

Process orders 100× faster

Your orders.
As data.
In seconds.

OCR reads any purchase order PDF and extracts supplier, date, product lines, amounts and VAT. No typing, no errors, and auditable boxes over every field it detects.

  • 99.4% accuracy
  • No templates needed
  • Data in the EU · GDPR

Constructora Ríos S.L.

Soluciones integrales en construcción

ORDEN DE COMPRA

N.º pedido: OC-2026-0427

Fecha: 2 de mayo de 2026


PROVEEDOR

Suministros Industriales del Norte S.A.

CIF: A-12345678

Polígono Industrial La Mina, Nave 7

08040 Barcelona, España

ENTREGA Y PAGO

Pago: 30 días F.F.

Dirección: Calle Mayor 12, Madrid

Divisa: EUR (€)

DESCRIPCIÓNCANT.IMPORTE
  • Cemento Portland CEM II 42,5RSKU-CMTC-425R1201 008,00 €
  • Acero corrugado B500S Ø12mmSKU-AC-B500S-12300345,00 €
  • Mortero de cemento M-7,5SKU-MOR-M7545171,00 €
  • Bloque hormigón 40×20×20SKU-BLQ-402020800656,00 €
Subtotal
2 180,00 €
IVA (21%)
457,80 €
Total
2 637,80 €

Documento procesado automáticamente · OCR

~10s
Average time to process a complete order, whatever the format.
99.4%
Extraction accuracy on structured fields.
0 templates
No need to set up each supplier's format one by one.
−90%
Manual typing time compared with traditional ERP workflows.

How it works

Three steps. Zero configuration.

No per-supplier templates, no rules to maintain, no scripts. Upload a PDF and get auditable data.

  1. 01

    Upload the PDF

    Drag it in or connect your mailbox. OCR supports native PDFs and scans. We process in batches.

  2. 02

    OCR extracts the data

    A model trained on real purchase orders. It detects header fields and every product line in seconds.

  3. 03

    Audit and export

    Check each field with its box over the original document. Export to Excel, CSV, JSON or your ERP via API.

Features

Everything you need to automate order entry.

Built for purchasing and accounting teams that receive dozens or hundreds of orders a month.

  • Full header

    Customer, tax ID, order number, date, delivery address, payment terms and currency.

  • Lines

    Line detection

    Every product: description, code/SKU, quantity, unit price, discount and amount. Multi-page tables included.

  • Auditable boxes

    Every extracted field has a bounding box over the original PDF. Review and correct in one click.

  • Verified calculations

    Automatic validation of subtotals, VAT and totals. If the numbers don't add up, we warn you before exporting.

  • AI

    Learns from your corrections

    Every time you adjust a field, OCR gets better. After a couple of orders from the same supplier, accuracy is practically perfect.

  • Export to Excel, CSV, JSON

    Clean format, ready for your accountant, bookkeeping or data warehouse. Structures compatible with SAP, Holded, Sage…

  • API

    API and webhooks

    Connect your system directly. Upload PDFs via API, get the result by webhook and forget the middleman.

  • Multi-currency and multilingual

    EUR, USD, GBP, MXN… Recognises documents in Spanish, English, Portuguese and French.

  • Data in the EU

    Servers in Germany, encryption at rest and in transit, GDPR compliant. Guaranteed deletion when you cancel.

01 · Extraction

From paper to table. In one click.

Upload a PDF (native or scanned, it makes no difference) and in under 10 seconds you get a structured object with every field. No rules, no templates, nothing to tweak.

  • Supports native PDFs and scans at 200 dpi
  • Batch processing (up to 500 PDFs at once)
  • Also ingests from a dedicated mailbox
  • Detects multi-page tables automatically

OCR

{"client": "Constructora…","cif": "A-12345678","orderNumber": "OC-2026-0427","date": "2026-05-02","currency": "EUR","lines": [{ "sku": …, "qty": 120 },{ "sku": …, "qty": 300 },… 4 items],"total": 2637.80}

02 · Audit

Every value, right where it is. Full traceability.

OCR draws a box over the original document showing exactly where each value came from. If you need to check or correct something, hover over it and edit in place.

  • Bounding boxes per field and per line
  • Inline editing in one click
  • Confidence per field (high · medium · review)
  • Change history for audits

Constructora Ríos S.L.

OC-2026-0427


2 de mayo de 2026

CIF: A-12345678

DESCRIPCIÓN · CANT · UNIT · TOTAL

  • Cemento CEM II 42,5R1 008,00 €
  • Acero corrugado Ø12mm345,00 €
  • Mortero M-7,5171,00 €
  • Bloque hormigón656,00 €
Total2 637,80 €

Confidence per field

Average98.8%

Ready to export

03 · Learning

The more you use it, the better it works.

Every time you correct a field, OCR remembers it for future orders from the same supplier or in the same format. It improves very quickly: from a supplier's 3rd or 4th document, accuracy is practically perfect.

  • Memory per supplier (tax ID)
  • No manual labelling needed
  • Your organisation's rules (ledger account, cost centre)
  • Learning shared across your whole team
Accuracy by number of orders from the supplier
100%90%80%70%86%1st95%2nd99%3rd100%4th100%5th

Orders processed from the same supplier

04 · Integration

Connect OCR to your ERP. Without waiting 6 months for IT.

Documented REST API, event webhooks and export to the formats your accounting team already uses. No installs, no per-user licences in your ERP.

  • Excel · CSV · JSON · XML
  • Connectors: SAP, Holded, Sage, A3, Odoo
  • REST API and webhooks per supplier
  • Zapier & Make included

OCRapi · webhooks

  • Excel
  • SAP
  • Holded
  • CSV
  • Sage
  • A3
  • Odoo
  • Zapier

Pricing that fits

A plan sized to your volume.

The price depends on the number of orders per month and the integrations you need. Tell us about your case and we'll design a plan that fits.

Frequently asked questions

What you're probably wondering.

  • You don't need templates. OCR uses a model trained on thousands of real purchase orders and recognises each supplier's format generically. If you receive orders with a very unusual layout, the model learns from your corrections after 2–3 documents.

  • It works with both. Native PDFs are slightly faster. For scans, we recommend at least 200 dpi and a document that's legible to the human eye. We also accept JPG and PNG.

  • Every field has a bounding box over the original document and a confidence score. Fields below 90% are highlighted in yellow so you can review them before exporting. You edit the value in place and OCR remembers it for the future.

  • Yes. We have native connectors for SAP, Holded, Sage, A3 and Odoo. If you use another ERP, we offer a documented REST API and webhooks. And for quick connections, Zapier and Make work wonderfully.

  • Your PDFs and extracted data are stored encrypted on servers in Frankfurt (EU). We don't use them to train generic models. If you cancel, we delete everything within 30 days. You have the rights of access, rectification and erasure under the GDPR.

  • You pay by the number of orders processed per month. Tell us how many PDFs you expect and which integrations you need, and we'll send you a tailored quote. There's a starter plan from 50 orders a month.

Stop typing orders. Today.

Send us 5 sample PDFs and we'll return the extracted data within 24 hours — so you can see the accuracy before committing to anything.

Already using OCR?