Bitewing is in private pilot with select DSOs. Request access.
Bitewing
Bitewing
§ Platform · E·01

The hard part is not the math.
It's normalizing ugly reality into something the math can trust.

Dental payment data arrives in an impossible mix of formats — 835 electronic remits, payer-portal exports, scanned PDFs uploaded to shared folders, faxes. Before you can audit anything, you have to normalize all of that into a structured dataset. That is the engineering wall. The Parser is Bitewing's answer.

Read-only ingestNo PHI leaves your tenantEvery field is citable
PARSER · INGEST STREAM
LIVE
INPUT
AETNA_835_20260312.txtParsed
cigna-portal-export-Q1.csvParsed
delta_eob_scan_2847.pdfParsed
united-fax-incoming.tiffParsed
OUTPUT
CODE
D2740
BILLED
$890
ALLOWED
$651
ADJ
CO-45
NET
Leased
TIN
94-2880556
DOS
2026-03-12
NPI
1538···247
CLM
0291-E
§ 01 · Inputs

Four walls the Parser breaks through.

Format fragmentation (835s, portals, PDFs, faxes). Contract opacity (~40% of practices do not have clean digital copies of their current contracts). Identity resolution (aliases, TPAs, TIN/NPI drift, schedule amendments). System fragmentation (a dozen PMS platforms, each with its own data model). The Parser reads all of them.

835 · ERA · portal · PDF · fax · OCR
835 / ERA
Clean, structured. The easy case.
Payer portal export
CSV/XLSX. Column names drift quarter to quarter.
Scanned EOB PDF
OCR + layout reconstruction.
Fax
Yes, still. Handled with the PDF pipeline.
§ 02 · Extraction

Every line item, every adjustment code, every footnote.

The Parser reads each EOB end-to-end — billed amount, allowed amount, adjustment reason code, routing footnote, TIN of the paying entity, date of service, provider NPI — and writes them as structured records with pointers back to the source document. When someone challenges a number, you can see the exact line it came from.

12,847 claims parsed this month · 99.4% field-level coverage (illustrative)
§ 03 · Normalization

The dozen PMS models collapse into one schema.

Once extracted, every field is normalized against a canonical schema: payer entity, TPA alias, contracted fee schedule reference, adjudication path, confidence tier. That is the structured substrate the Variance Router and the Contract Engine run on. Nothing downstream has to guess.

canonical TPA alias map · versioned fee schedules · provider/location/TIN graph
PMS A
PMS B
PMS C
Canonical schema · one table to reason over
Book a walkthrough

First variance report, target: 14 days.

Read-only access to one location. Two weeks. Priced, cited, appealable variance — or a clean bill of health.

Book a walkthrough