Last updated: 3 July 2026. This English version is authoritative; a Chinese version is also available.
1. Who we are
Zen Horizon (zh.app, Chinese name 纵横研报) is an equity research publication operated by H&A Global Limited, a company registered in England and Wales. We publish long-form research reports on AI, technology and global equities from a long-term business-owner's perspective, in up to nine languages. Zen Horizon is an independent publication; it is unrelated to the Horizen (ZEN) cryptocurrency project or to any brokerage. Operator details and contact information are on the About & Contact page.
2. How a report is produced
Each report goes through a fixed pipeline before publication:
- Research brief. Every report starts from a structured task card that fixes the company, the analytical framework (Section 3) and any reader-requested questions.
- AI-assisted deep research. Drafts are generated with AI deep-research tooling over public sources — filings, earnings materials, industry data and reputable press.
- Automated verification gates. Before a draft can enter the database it must pass hard checks: number-fidelity digests (figures may not silently change through any rewrite step), structural checks against the framework template, and coverage checks that reader-requested questions are actually answered. A failed gate blocks publication.
- Editorial review. Reports are human-reviewed before going live. Readability rewrites are constrained by the same number-fidelity gates: wording may change, figures, ratings and conclusions may not.
3. Research frameworks
Reports follow one of three published frameworks, chosen per company:
- Zen Horizon framework — our default lens: what a long-term part-owner of the business would need to know about the industry chain, competitive position, unit economics and valuation.
- Buffett-style framework — value-oriented questions for durable-moat, cash-generative businesses.
- SUN-R framework — for high-volatility, narrative-driven names (crypto-adjacent platforms, structural-shift plays): structural migration, real usage, narrative liquidity and zero-risk assessment.
4. Ratings, fair-buy price and valuation band
- Rating. Each report closes with a rating on a ladder from Strong Buy through Buy, Cautious Buy, Hold/Neutral and Watch down to Underweight/Sell. The rating is the analyst conclusion of that report at its publication date; it is never updated silently — a changed view requires a new report.
- Fair-buy price. Where given, the fair-buy price is the upper edge of the range in which we would consider the stock attractively valued. Above it, the report's thesis relies on growth already being priced in.
- Valuation band. Report pages show the fair-buy anchor against the live market price so readers can see valuation position at a glance. Live prices come from an automated daily feed with currency-consistency checks.
5. Baillie growth scorecard
Most covered companies carry a ten-question growth scorecard inspired by long-horizon growth investing. Each question is scored 0–10 against the report's verified evidence, giving a 0–100 total. Load-bearing figures (market cap, margins, market share) are re-checked against current public sources at scoring time. Scoring is deliberately conservative: where evidence is mixed, the lower score wins. The Growth Board ranks all scored companies.
6. Languages and translation fidelity
Reports are written in English or Chinese and translated into the other supported languages. Translations pass automated fidelity gates: the full multiset of figures must match the source (including magnitude words such as "tens of billions"), and ratings may never flip in translation. Where a translation does not yet exist, we show the source language rather than machine-fallback, and the page says so.
7. Freshness and re-research
Published reports are dated and never silently rewritten. A staleness monitor tracks price drift against each report's valuation anchor, report age and earnings events; when a report goes stale we commission a new report with a new URL. The older report stays online for the record, with a visible notice linking to the newer one.
8. Corrections
Factual errors are corrected in the published report, and the page's "updated" timestamp changes. Corrections that affect a rating or conclusion are handled as re-research (Section 7), not as silent edits. To report an error, email contact@zh.app with the report URL; we aim to respond within two business days.
9. Independence and use of AI
- Part of our coverage is initiated by reader research requests (a membership benefit). Readers choose what we cover; they have no influence on ratings or conclusions, which are never for sale.
- We disclose AI assistance site-wide: research drafts are AI-generated and pass the automated gates and editorial review described above before publication.
- Nothing on zh.app is investment advice. All content is general research commentary for informational and educational purposes — see the Terms of Service.