How this publication is made
The idea
Every finance product has more data than this one. None of them let you check their numbers. Equal is built on a single trade: cover fewer companies, but make every figure traceable; to a filed statement, a regulator’s publication, or a stated, adjustable assumption. The name is the thesis: the reader should stand on equal footing with the analyst.
Built by one person
Equal is designed, researched, written and engineered solo: every narrative hand-authored in two languages, every bank figure hand-extracted from OJK publications, every pipeline written for one maintainer. Constraints shaped the architecture: deterministic numbers cost nothing to serve, curated analysis beats generated text, and the only AI in the product is a cached, citation-checked trend memo.
The standards
- Accuracy first, zero tolerance: a partially verified claim presented as complete is treated as an error.
- Append-only data: published figures are never silently regenerated underneath the narratives that cite them.
- Analyses take a position and state what would prove it wrong (the codified sell-side standard in the repository).
- Both languages are first-class: the Indonesian text is an analysis, not a translation afterthought.
- Honest gaps: em-dashes, NM markers, pen-up charts, stated limitations.
The stack
A TypeScript monorepo: a shared deterministic engine (ratios, banks, valuation, DCF; every function unit-tested against hand-computed fixtures), a Fastify API serving committed data snapshots, and a Next.js front end. No chart libraries, every chart is hand-rolled SVG. No component libraries, the editorial look (Fraunces, Cormorant Garamond, Lora, Space Mono on warm paper) is the design system.
Read next
Educational analysis, not investment advice.