TwoMargins: Crowdsourcing Financial Document Analysis for Wall Street

The coolest, most constructive best-yet crowdsourcing effort to hit Wall Street

-- Wall Street Journal

Historical engineering work

TwoMargins was built to change how financial information could be analyzed collaboratively. The product is presented here as historical engineering work; current availability is not implied. Its public record includes Wall Street Journal coverage and Bloomberg's archived use of its annotation layer for Warren Buffett's 2015 shareholder letter.

The platform allowed analysts, investors and researchers to annotate and discuss SEC filings and other financial documents collaboratively. The concept sounds simple: ingest financial documents and let people discuss specific passages. In practice, the parsing and anchoring problems were difficult.

Bloomberg and the Buffett letter

When Warren Buffett published his 2015 annual letter to Berkshire Hathaway shareholders, Bloomberg chose TwoMargins as the platform for their annotated interactive edition. Luke Kawa at Bloomberg Business wrote the annotations, and readers could follow along, highlight passages, and add their own commentary directly on the document through TwoMargins' annotation layer.

This was one of the platform's highest-profile public uses. The shareholder letter was converted into annotatable HTML through the document-processing pipeline while preserving the financial tables and document structure needed for reading and discussion.

The hard problem: parsing SEC documents

SEC filings are not clean, well-structured HTML. They are dense, deeply nested documents with irregular formatting that varies between companies, filing types, and time periods. Building a system that could reliably parse these documents, preserve their structure, and overlay an annotation interface on top required solving a genuinely difficult engineering problem.

We built a Nokogiri-powered HTML parsing system specifically designed for this challenge. The parser needed to:

• Handle deeply and irregularly nested HTML without breaking document structure
• Preserve the original formatting so annotations could be anchored to specific passages
• Support documents ranging from a few pages to hundreds of pages in length
• Process new filings quickly when they became available (some document releases are time-sensitive market events)

The Buffett letter was a perfect example of the complexity: a 30-page PDF converted via BCL easyConverter SDK into HTML with 30 distinct page divs, each containing multiple nested content blocks, table layouts for financial data, and precise positional CSS. Our parser had to ingest this structure, normalize it for the annotation engine, and serve it through a real-time collaborative layer — all without breaking any of the formatting that made the financial tables readable.

What we built

Document ingestion pipeline: We built an automated system that gathered data from SEC's EDGAR (Electronic Data Gathering, Analysis, and Retrieval) system. New filings were detected, downloaded, parsed, and made available on the platform with minimal delay. For user-uploaded documents like the Buffett letter, the pipeline handled arbitrary HTML structures with the same reliability.

Annotation engine: The core feature — the ability to select any passage in a financial document and attach annotations, comments, and analysis. Built on the Annotator.js library with a custom Store plugin, the system mapped text selections to document positions, persisted them via a Rails JSON API, and rendered highlighted passages with annotation counts. The engine supported threaded replies, so annotations became discussions.

Collaborative features: TwoMargins was built for group analysis. Multiple users could annotate the same document, respond to each other's annotations, and build shared understanding of complex filings. Social authentication (Twitter, Google, StockTwits) lowered the barrier to participation, and guest posting allowed anonymous contributions.

Embeddable annotation layer: The platform supported oEmbed and iframely standards, so the annotation interface could be embedded in third-party sites — which is exactly how Bloomberg used it. Their feature article linked directly to the TwoMargins-hosted annotated document.

Supporting a live product

Financial markets don't wait. Some of the most valuable moments for TwoMargins were during high-profile document releases — earnings reports, regulatory filings during U.S. presidential elections, and other market-moving publications. The Bloomberg Buffett letter was one of the biggest: a featured article on Bloomberg.com driving traffic to an annotation layer we hosted and maintained. We supported the product team during these peak traffic events, ensuring the platform remained responsive when user activity spiked.

Results

TwoMargins was featured in the Wall Street Journal as "the coolest, most constructive best-yet crowdsourcing effort to hit Wall Street." Bloomberg chose the platform for their marquee annotated Buffett letter. The platform attracted a community of financial analysts and investors who used it to collaboratively analyze some of the most important financial documents in the public markets.

Our role

We were responsible for the full technical build: the document parsing pipeline, the annotation engine, the collaborative features, server infrastructure, and deployment. As the project matured, we also onboarded and trained new developers to ensure the team could continue building without us as a bottleneck.