Opportunities › Structured data extraction from unstructured documents › Proposed app
Sourcemark
Structured data from documents, with a cited source for every value
- Score
- 37
- confidence 49
- Evidence
- 6
- items on the page
- Rivals
- 2
- named in sources
- First version
- 45 d
- estimate
- Type
- Desktop app
- PWA: runs in the browser, installs on Mac and Windows
- Market since
- 2017
- Established market · AI estimate
ScreensConcept
Desktop app: Teams extract data from document piles, review citations and manage quality control in a workspace with dense schemas and verification tasks. Sample screens on a computer, in the proposed brand. Their numbers are sample data.
Schema: Claim file
Rules: 3 checks12 fields
About this appAI proposal
Sourcemark turns piles of PDFs, emails, transcripts and spreadsheets into structured records that match a schema you define. Every value comes with a citation to the exact words it was read from, even when the answer is spread across several documents.
It is built for analysts and technical teams in finance, insurance and other regulated work. The engine searches all the documents for every relevant value instead of sampling. Where sources disagree or evidence is thin, it flags the field so a business expert can review it.
Use the API for volume, or the review workspace for checking. Reviewers see the value beside the highlighted source, accept or correct it, and the decision is logged. The result is data you can defend to an auditor.
- Tagline
- Every value, traced to the words it came from.
- Voice
- precisecalmplainspoken
- Why this name
- A source mark is the small note that ties a claim to its origin. The name says the product's one promise: each value points back to its source.
In the first version
Members say which features they would use. It tells PWAS what people want; it changes no score.
Schema-first extraction
Define fields, types and rules once. Output always matches the schema, ready for databases, spreadsheets or downstream code.
No member yet
Word-level citations
Each value links to the exact words and page it came from, including values built from several sources.
No member yet
Cross-document reasoning
Combine, reconcile and check values scattered across files, such as a policy, its endorsements and related emails.
No member yet
Exhaustive search
Looks through every document for all matching values rather than sampling, and reports what it did not find.
No member yet
Brand identityAI proposal
Check the name against trademark registers before using it.
Palette
#1F4D3FLedger Green · primary#F6F3EAPaper · surface#1B1F1DInk · text#F2C94CHighlighter · accent, citation marks#B3392BFlag Red · uncertainty flag#5F6B66Slate · secondary text
Typography
Aa
Sourcemark
PricingAI proposal
The material gives no competitor prices. These tiers are a starting point to test, based on the suggested volume and seat pricing in the opportunity.
Trial
Free
- 100 pages a month
- One schema
- Citations and review view
Team
$299 a month
- 5,000 pages a month
- Unlimited schemas
- 3 reviewer seats
- Webhooks and CSV export
Business
$1200 a month
- 30,000 pages a month
- 10 reviewer seats
- Lineage export packages
- Audit log and role controls
Enterprise
What it could be worthEstimate
What a software business making this much sells for today, from real deals. Pick a monthly revenue.
If it makes
Worth about
$194,000
Sale range $132,000 to $263,000
$120,000 a year in revenue
From 87 software businesses listed and sold on Empire Flippers in the last year, revenue vetted by the broker: they sell for about 19× their monthly revenue, most between 13× and 26×.
Measured on profit, the middle deal is 32× monthly net profit: about 2.7 years of profit.
An estimate, not an offer or financial advice. A real price depends on profit, growth, churn, how much of the work is the owner's, and the buyer.
How it would growAI proposal
How the product would reach new customers through its own use, and how the first ones would be found. A plan to test, not a result.
- Built-in loop
- Shared outputs. Reviewers share read-only evidence links showing a value beside its highlighted source passage. Auditors, colleagues and clients open them to confirm a figure without the whole dataset.
- Do they become customers?
- Link recipients are often auditors or managers who could pay for seats, but many only view once. Test how many shared evidence links lead to a trial or a reviewer-seat request.
- First customers
- Small insurance claims and underwriting operations teams, plus finance analysts at mid-size lenders, who now pull fields from document bundles by hand.
- Where to reach them
- Hacker News and dev forums: Share a demo that extracts a schema from a messy public document set, with citations shown, and invite technical teams to try the API.
- LinkedIn outreach: Message operations leads in insurance and lending with a free pilot on one of their document bundles.
- Compliance communities: Write about verifiable extraction and data lineage for audit teams, with an example evidence package.
- First milestone
- Five teams in one vertical each run a real document batch, and at least three pay for a monthly plan.
Why people already pay for this
- VerifiedHacker News post received 56 points and 54 comments
- Public claimOrganizations attempt to use LLMs like Claude for unstructured data extraction but hit system limitations around file volume, input type, cost, and latency.
- Public claimThere is demand for extracting information from scattered sources that requires reasoning across documents rather than point-by-point extraction.
- Public claimA core friction point in adoption of data extraction solutions is the inability to quickly verify and trust results.
Tech stack
Every proposal builds on the same PWA base, installable from the browser on any phone or computer. Dashed boxes are what this app adds.
- AppNext.js and TypeScript
- PWAManifest and service worker
- UITailwind CSS
- Data and authSupabase Postgres
- PaymentsStripe
- EmailResend
- HostingVercel
- This app · Document processingOCR and multimodal model APIReads scans, tables and page images, and returns word positions for citations.
- This app · File storageObject storage with encryptionHolds large document batches and source files for the highlighted view.
RisksAI inference
- AIModels may still misread scans or invent values. Citations and flags help, but a wrong value with a plausible citation could pass review. Accuracy must be measured per field.
- CommercialRegulated buyers buy slowly and want security reviews early. Teams may also keep using general LLMs with prompts if the gap in verification does not feel big enough.
- TechnicalExhaustive search across large document sets is costly in compute and time. Pricing per page may not cover model costs on long or image-heavy files.
- RegulatoryInsurance and finance documents contain personal and sensitive data. Buyers may require data residency, retention limits and certifications before a pilot.
- PlatformThe product depends on third-party model providers whose prices, limits and terms can change, which affects margins and reliability.
Everything marked AI proposal or AI inference was written by a model from the evidence on the opportunity page. The proposed name must be checked against trademark registers before it is used.