Skip to main content
DudoxxSentinel
Sign in

Full-service data delivery

Your paper backlog, delivered as a checked dataset

You send the pages. We deliver data you can use, with every doubtful field checked by a person.

A data manager lifts a stack of filled-in paper forms from an archive box beside a document scanner, with the Hamburg skyline behind the window.

What it is

We process your paper backlog end to end: intake, reading, review of every doubtful field and export. You do not tune or operate anything. The same holds for letters, reports and recordings: we deliver a knowledge bundle per document, with every passage tied to its place on the page.

Who it is for

Registries, trial units and institutes that hold a backlog of paper forms, faxes or questionnaires and have no capacity to key them in.

What you get

  • A checked dataset in your target structure, as CSV or Excel
  • A codebook that explains every column and code
  • An audit file: who changed what, when, and the original value
  • The image crop of every field a person checked

How it runs

  1. We agree the target structure and the forms in scope with you.

  2. You send the pages. We take them in and tie each page to its document.

  3. The pages are read. Fields the checks do not agree on go to a person with their image crop.

  4. We export the checked dataset with codebook and audit file.

Diagram: paper backlog (scans, fax, photos), intake and sorting, reading by the Dudoxx pipeline, then certain fields are accepted and doubtful fields go to a person with the image crop, ending in a checked dataset as CSV or Excel with codebook and audit file.
From paper backlog to checked dataset.

What we need from you

  • A sample of your forms and one blank copy of each form type.

  • Your target structure, or a short call to design it with us.

  • A data processing agreement, signed before any real document is shared.

Typical questions

Where does our data go?
Sentinel runs on Dudoxx servers in Germany or on your own server. No outside AI provider sees your pages, and your data trains no model. A data processing agreement is signed before any real document is shared.
What if a value is wrong?
Doubtful fields do not enter the dataset as guesses: they go to a person with the image crop. Every change shows who, when and the original value, and a re-run never overwrites a human correction. In the latest run on synthetic test forms, 1 wrong value was kept in 1,272 fields.
How long does it take?
It depends on your volume and on how many fields are flagged for review. A test on your own forms gives the real figure before you commit.

Tell us about your forms

A short call is enough to see which service line fits. Or look at how we measure first.