Build products from XBRL data.

XBRL consulting traditionally means helping companies tag their SEC filings with XBRL markup for regulatory compliance. We do something different: we help companies build products from XBRL data. We help AI teams, data engineers, and fintech companies extract, normalize, and build on the structured financial data embedded in SEC filings — not file it.

XBRL provides structured, machine-readable financial data tagged with standardized taxonomy concepts. Research shows that using XBRL instead of raw HTML reduces financial extraction errors by 74× — from 8.16% to 0.11%. For AI applications, this means dramatically fewer hallucinations when answering financial questions, because every number is tagged with its precise semantic meaning rather than scraped from ambiguous HTML tables.

HTML SEC filings are formatted for human reading — tables, paragraphs, and formatting. XBRL (specifically Inline XBRL or iXBRL) embeds machine-readable tags within the HTML. When you parse HTML, you're guessing what numbers mean. When you parse XBRL, the filing tells you explicitly: this number is Revenue, tagged as us-gaap:Revenues, for fiscal year 2024, in USD. That semantic precision is what makes structured extraction reliable at scale.

The most common issues: companies using custom taxonomy extensions instead of standard US-GAAP concepts (making cross-company comparison difficult), inconsistent tagging of the same line item across years, incorrect negative signs, dimensional structures that vary between filers, and taxonomy version changes. According to XBRL US data quality research, these affect roughly 15–20% of filings and require deep domain expertise to handle correctly at scale.

Yes — that's one of our core deliverables in the Pipeline Build engagement. We build production-ready APIs that extract normalized financial data from XBRL filings, handle taxonomy mapping, resolve custom extensions, and serve clean JSON endpoints. The API typically covers income statements, balance sheets, cash flows, and custom data points specific to your use case.

A working prototype takes 1–3 days in our Sprint engagement. Production-grade with full taxonomy mapping, error handling, and multi-company normalization: 2–4 weeks in our Pipeline Build. Compare this to building in-house, where teams typically spend 6–12 months learning EDGAR quirks before achieving reliable extraction.

We build on edgartools — our own open-source Python library with 5M+ downloads, nearly 1M monthly downloads, and 2,000+ GitHub stars. It handles EDGAR access, filing retrieval, XBRL parsing, and financial statement extraction. For production, we integrate with PostgreSQL, Redis, and your cloud infrastructure (AWS, GCP, or Azure). For AI agent teams, delivery is via MCP server integration.

Data APIs give you raw access. Consulting builds the product layer on top. If you need custom taxonomy mapping, cross-company normalization beyond what APIs provide, XBRL-to-LLM integration, or a proprietary pipeline you own and control — that's the gap. Many clients start with a data API and come to us when they hit its limits.