docRefinery vs DocuPipe

Two answers to layout variety

DocuPipe's headline is consistent data from documents that never look the same. You describe the value you need, not where a template puts it, and one schema handles every layout. docRefinery chose the other way on purpose. It recognises a sender's layout, learns it once and recalls it each time that sender's documents return. If every document is new to you, DocuPipe's approach is the natural fit. If the same few hundred vendors send most of your volume, ours is.

Beyond that, the products are close. Both classify and split mixed packets, score each field, draw values back onto the page and route uncertain ones to a person. Anyone comparing them should know that before reading further.

Where DocuPipe is the better choice

• Certifications in hand. DocuPipe says it is ISO 27001 certified and HIPAA and GDPR compliant, and can sign HIPAA Business Associate Agreements. For healthcare and insurance buyers that can settle the decision. docRefinery's SOC 2 Type II audit is in progress with no report yet, and we hold no ISO 27001 certificate.
• Documents that do not repeat. If each file is from a new sender, a layout remembered on the first visit is never used again, and a schema-first tool is simpler.
• A public benchmark. DocuPipe publishes a 97.02% field accuracy figure on 72 hard public documents and has released the benchmark as open source. We publish no accuracy figure.
• Self-serve paid plans. Business starts at $99 a month with included credits, and the docs list credit costs per step: parsing plus standardising costs 3 credits a page. Our paid rates come from sales.
• Private deployment that is well documented. DocuPipe installs into your own AWS, Azure or Google Cloud account, or onto Kubernetes, including OpenShift and air-gapped clusters. It is an enterprise agreement with a minimum annual credit purchase.

What docRefinery does that DocuPipe does not, or does differently

The products share a lot, so the differences are specific. These are the ones that decide it.

• Layout memory. Once a layout is remembered, the second document from a vendor is recognised on sight, and over time review effort falls as fewer values drop below their thresholds. When a vendor changes its template, the new layout is reviewed rather than silently misread: see vendor layout changes.
• You define the output. Each field has a plain-language instruction, a "Return as" name you choose as its output key, a type and a required flag, governed by your organisation rather than supplied by each integration.
• One output format for every document. Every field returns text, alias, field_confidence and location, and every table returns rows, so one parser handles bank statements, utility bills and IDs alike.
• Governed fields. Owner, admin and member roles decide what may be extracted, and a caller cannot override it.
• Null, never a guess. A field missing from the page comes back null, and a score can be lowered by review and never talked up.
• Corrections that become fixes. A reviewer's correction becomes a suggested fix, tested on that layout's own documents before anyone applies it.
• Rules, versions and replay. Business rules run on every document, and each change is a version you can replay against past runs before it goes live.
• Custom types at no charge. Document types outside our catalogue are trained, hosted and re-trained at no charge, and you pay only for extraction volume.
• Plug and play. API keys with expiry, revocation and an activity log, export to JSON, CSV or XLSX that always reflects corrections, and a Chrome extension in early access that fills web forms from a document.