AdaptivMapr
Product
Schema mappingStructural reshapeAny-to-any convertTemplates & packsValidatorsFHIR R4 emitSchema-only vs full-dataLayout reuse & driftAll product
Solutions
HealthcareFinance & paymentsHR & peopleBusiness packsBy the job you haveAll solutions
Developers
API referenceQuickstartMCP serverConnectors & webhooksIntegrationsCompliance & trustChangelogStatusBlogAboutAll developers
Pricing
Get an API keyLog in

Schema mapping for regulated data

Any file you are sent, mapped to your schema.

Most columns are settled before any model is consulted. Edit the header row — it runs the real cascade, and nothing but the column names leaves your browser.

Start mappingSee how it works
Personalnummeremployee_id1.00 · heuristic
Namelegal_name0.85 · heuristic
E-Mailemail1.00 · heuristic
AHV-Nrnational_id0.85 · heuristic
IBANiban1.00 · heuristic
Eintrittsdatumhire_date1.00 · heuristic
6 of 6 resolved · no AI callSchema-only · nothing stored
VornameNom de familledata_di_nascitaAHV-Nr.E-Mail AdresseRechnungsnr.

This is what actually arrives.

Fällig amn° clientCODICE FISCALEGeschlechtIBAN / BICEintrittsdatum

Not one of them is first_name. The catalogue carries hints in German, French, Italian, Spanish and English, which is why most of these are settled before anything is metered.

How it works

Four steps, and you see the reasoning at every one.

  1. 01

    Upload

    Send the file, or point us at a connector. In schema-only mode the headers and three clamped rows are all that leave your network.

  2. 02

    Match

    The cascade proposes a field per column. Cheapest layer first, and it tells you which layer decided and how confident it was.

  3. 03

    Review

    You correct what is wrong. A medium- or high-risk template flags itself and will not commit without an explicit approval.

  4. 04

    Commit

    Rows out in your shape. The corrections are learned for next time and the whole mapping is written to the audit log.

One engine, six jobs

Everything between a file
and a row you trust.

Most import tools do the easy half — a clean CSV with the right headers. These six all run on the same engine, with the same audit trail.

Map a messy file

Column names nobody normalised, in five languages, with typos — matched to your fields and scored, so you can see which ones to check.

Read a scan or a photo

A faxed lab report, a photographed form, a PDF with a real table in it. OCR returns the grid, and the grid feeds the same mapper.

Rebuild a workbook

Transpose, pivot, join across sheets — and hand back your own template with its macros intact.

Never map it twice

The confirmed mapping is cached against a fingerprint of the header row. The same sender next month replays it in one click.

Notice when they change it

A known sender arriving with a new field set is flagged as drift, with the added and removed columns named.

Check it against the standard

IBAN and BIC by checksum, not by regex shape. LOINC, ICD-10, ATC, CPT, NPI, GTIN — field by field.

What leaves your building

The guardrails are not settings someone forgets.

Schema-only mode sends your column names and at most three sample rows, each truncated to 80 characters — enforced at the HTTP edge and again at the model boundary, by one function, so the two halves cannot drift apart.

The compliance posture
Ten rows stay inside; a header row and three rows clamped to 80 characters are all that crossyour networkwhat leaves………≤ 3 rows · ≤ 80 charsclampForSchemaOnly()
13 of 33 templates gate on review7 packsCH · EU · US residencyErasure is one implementation

Any format in

The unstructured path is the same path.

CSV and TSV, Excel — including macro-enabled .xlsm and .xltm workbooks, read but never executed — JSON and NDJSON, XML, SQL over HTTP, Parquet, a Word document, a scanned PDF, a photograph of a form, a dictated handover, or a URL.

Everything that is not already a grid is turned into one first: OCR returns a real table and the table feeds the same mapper, so a faxed lab report and a clean export are resolved by the same cascade, checked by the same validators, and written to the same audit log. There is no second-class input.

The convert path
Nine input formats converging on one schemaCSVXLSXXLSMJSONXMLSQLPDFIMAGEAUDIOyour schema
7 template packs, sized by how many templates each carriesHealth10CRM5Commerce5Finance4HR4Core3Evidence2

The catalogue

33 templates, already written.

Across 7 packs — healthcare, finance and payments, HR and people, plus core business, CRM, e-commerce and clinical evidence. Each template carries field-level validators (IBAN and BIC by checksum, LOINC, ICD-10, ATC, CPT, NPI, GLN, GTIN), multilingual hints in German, French, Italian, Spanish and English, and a FHIR R4 resource mapping where one applies.

13 of them are medium or high risk and gate on human review before they commit — derived from each template's own risk field, so the gate cannot drift from the catalogue. Bring your own schema instead and the same cascade maps onto it.

Browse the catalogueSee one template

Every map, deterministic or not

$0.001/ map

About a thousand maps for a dollar. A re-map of a header row you have confirmed before costs the same and uses no AI at all.

$10 prepaid walletor $6/monthno seats · no contractPHI +20% under a BAA
See what a map costs

Questions

What people ask first.

The awkward ones, answered in full: what a map costs, what leaves your network, what happens to a file after it is mapped, and which compliance claims we will not make.

Compliance in detail
How is schema-only mode priced?
There is no free tier. You top up a prepaid token wallet (from $10, shared across our suite) and every map draws a small flat fee — a few tokens — including a fully deterministic map and a cache hit that used no AI. Schema-only is a data-minimization mode, not a price tier: only your column headers and up to 3 clamped sample rows (≤80 chars each) are processed. A DPA/BAA may still apply depending on the data; we provide one on request.
What does PHI routing cost, and what unlocks it?
PHI/enterprise routing adds 20% to the whole map charge, flat fee included, and it is locked until your workspace accepts the BAA/NDA in-app under Settings → Security & Data. Standard routing is what a new workspace gets by default and pays no uplift. The surcharge follows the routing a run actually received, never the policy you set.
Does AdaptivMapr store my regulated data?
In schema-only mode we store nothing beyond the clamped samples — headers plus up to 3 truncated rows; raw records never leave you. The cascade engine itself is stateless pure compute. Uploads live in Cloudflare KV under the retention window your workspace picks — destroyed at commit, or 24 hours (the default), 7 days or 30 days, quoted back on each upload as expires_at — and account metadata and audit logs persist under per-tenant isolation. In full-data mode the payload is routed to a PHI-eligible provider in your region for the model step and is not retained by us.
Do you replace Flatfile or Redox?
No. We are not a generic CSV importer and not a full integration engine. AdaptivMapr is a self-serve schema-mapping layer for regulated data — healthcare, finance and PII — that maps CSV, Excel, SQL and JSON to your schema, and emits FHIR R4 where the template carries a resource mapping, without the cost or the calendar of a heavyweight integration project.
Are you HIPAA certified?
HIPAA is not a certification anyone can hold, so nobody is. What we offer is a BAA, a HIPAA security risk assessment (SRA) on file, GDPR-friendly handling, and SOC 2 in progress. We say "HIPAA-ready", never "HIPAA certified".

One key, from a messy file to your schema.

Top up a $10 prepaid wallet and start mapping. In schema-only mode only your headers and three clamped rows ever leave you.

Get an API keyRead the docs