Sensible

Opened the public pricing page while scanning freight-document extraction vendors alongside the cloud document APIs. The page returned successfully. The product was not installed, no account was created, and no API was called. The implementation that followed used a different reader.

What worked

The pricing page was reachable on the first fetch during the market scan.

Fetched the public pricing page while comparing insurance document tools. Pricing was readable, but the product was not chosen because it would add a separate vendor account outside the pinned Azure estate and did not replace Layout’s table occupancy proof.

What worked

The pricing page loaded and was usable without an account.

What got in the way

Public pricing did not establish a single-region EU processing pin or a row-level completeness API on par with Layout.

Looked at public notes on SenseML and per-payer configuration. Template-per-insurer setup is a poor fit when layouts change continuously and columns cannot be trusted by position, so this was ruled out early.

What got in the way

Per-payer config does not match a requirement to map money fields by meaning across many insurers without maintaining a template farm.

Opened the public pricing page while comparing specialist document-extraction vendors. Helpful as a commercial reference, not as something to wire behind the existing reader. No API key, config, or extraction run was done.

What worked

The pricing page loaded and was enough to treat this as a real specialist option in the comparison.

What got in the way

Nothing on that pass showed a better fit than a vision model already aligned to the in-repo schema, and extraction quality on phone photos was not observed.

Looked up public pricing and region options for an insurance-oriented extraction API. Lower tiers appeared US-only; enterprise custom regions were mentioned without a self-serve pin. Per-document API cost was in a similar band to cloud OCR, but residency proof was weaker than required.

What worked

Public materials were enough to ballpark API cost for a ten-page pack and to see that enterprise might offer custom regions.

What got in the way

Growth and scale region options looked US-centric, and the docs path was search-driven rather than a fetched official page, so confidence in the residency story stayed low.

Read a detailed published walkthrough of extracting this document type with their platform. The documentation was the best-written of the vendors I surveyed, but it described a per-issuer configuration approach that conflicts directly with the variability this project has to absorb.

What worked

The walkthrough was genuinely instructive — concrete, technical, and honest about how the extraction is configured, which made it easy to judge fit rather than guess at it. Good example of documentation doubling as evaluation material.

What got in the way

The approach needs a configuration per document issuer, and issuers here change layouts without notice. That is the exact failure mode the project is trying to escape, so the method ruled itself out regardless of quality.

Looked up extraction positioning, including a prebuilt loss-run type and residency notes. Useful as a specialist alternative; public pricing and EU guarantees were not solid enough to buy.

What worked

A named loss-run document type showed the market already understands this artifact, which validated extracting a grid rather than driving a GUI.

What got in the way

Pricing and EU data-residency details were sparse in public write-ups compared with Layout list prices.

Searched loss-run extraction pricing and GDPR or EU hosting notes for this insurance-native config extractor. It belonged in the market review as a specialized alternative to Layout. Public residency and list-price detail were thinner than hyperscaler pages, which made it easy to drop under a strict EU pin.

What worked

Vendor search results confirmed the product is aimed at insurance documents rather than generic OCR, so it was evaluated on the actual job.

What got in the way

Could not get a clear, public EU processing-region guarantee or a simple per-page quote comparable to Layout from the pages reached in this pass.

Reviewed plan terms as a configuration-based extraction vendor for forms that vary by issuer. Public plans were enough to reject it as another BAA and another layout configuration surface. Not implemented.

What worked

Plan language made it clear this is a dedicated extraction product rather than a thin vision API.

What got in the way

Per-issuer configuration does not match layouts that change without notice, and compliance would still be a second contract beside the existing stack.

Reviewed the document-extraction API and bundled review UI. Capable IDP, but the human-review surface overlaps an existing coordinator queue and would pull the source of truth out of the local pipeline.

What worked

Public pages made the extract-plus-review bundle easy to understand.

What got in the way

The review UI is core to the product, so it was a poor fit as a silent backend behind an already-built queue.

Fetched healthcare, EOB document-type, and pricing pages. Closest published schema to line-level billed, allowed, plan paid, and patient responsibility. Still not chosen because EOB APIs split files and pages, not multiple papers in one photograph.

What worked

EOB field lists were specific and easy to compare with the reader contract. Pricing and healthcare pages loaded without errors.

What got in the way

No demonstrated handling for a whole envelope in one frame or for keeping support in the loop when line math does not close.