Methodology
How we source, structure, and verify the data behind every Insight report — written plainly so you can decide whether to trust it.
674
published reports
8,861
cited sources
1,140
Market profiles
Numbers recompute on every build from the underlying source packs.
Source families
Five families feed every report. Most claims rest on the first three; commercial aggregators and trade press are corroborators, never sole anchors.
Thai government + regulator
Bank of Thailand (PromptPay, e-payment series, licensee registers), Board of Investment, Thai Customs, Ministry of Commerce trade stats, ETDA surveys, SEC Thailand. Graded primary.
Company filings (listed + private)
SET 56-1 One Reports, SEC EDGAR 20-F filings, and — crucially — Department of Business Development (DBD) annual financials for private Thai entities (TikTok Shop, Flash Express, LMWN, Ascend Money). Graded primary.
Multilateral + international body
UN Comtrade, FAO FishStat, NOAA Fisheries, EU DG MARE, EUMOFA, World Bank, IMF, OECD. Graded well-established — counts at full weight for market-sizing, demand, and recency gates.
Trade bodies + rule text
Federation of Thai Industries, TTIA (Thai Tuna), TIPMSE (packaging EPR), TISI standards, Royal Gazette rule text, draft legislation from ministry consultation portals. Primary for regulation; supporting for sector aggregates.
Commercial aggregators + trade press
Mordor Intelligence, Grand View Research, Euromonitor, Bangkok Post, Nation Thailand, SeaFoodSource. Graded supporting. Used for corroboration and colour, never as the sole anchor.
Process
Seven steps from scoped market question to published deliverable. Each step has its own canonical artifact so later refreshes reuse the same inputs.
STEP 01 · INTAKE
Every report starts with a scoped market question, target source families, and an explicit evidence standard before drafting begins.
STEP 02 · SOURCE PACK
Sources are collected into a traceable pack with provenance, access date, extraction method, and visible freshness/evidence grading.
STEP 03 · CLAIMS
Report claims are separated from raw source text so every commercial assertion can be traced back to specific evidence.
STEP 04 · NORMALIZE
Key stats, benchmark rows, entities, charts, and source metadata are normalized into reusable structured inputs instead of one-off report copy.
STEP 05 · REVIEW
Analyst review adds context, confidence scoring, conflict notes, and limitation handling. Quantitative claims stay tied to specific sources.
STEP 06 · DELIVER
The premium web report is the default buyer-facing output. CSV, PDF, data-pack, and source-appendix files are generated only when scoped or already present.
STEP 07 · PUBLISH
Finished reports go live with a defined update cadence. Later revisions reuse the same canonical inputs instead of rebuilding the product by hand.
Quality control
What sits between the source pack and the published report. Each control has a defined owner and runs on every report, not just the next one.
Automated ingestion
Sources are pulled, parsed, and normalized through a versioned pipeline so every claim links back to a snapshot, not a screenshot.
Cross-source matching
Conflict checks run across overlapping anchors — for example, BOT survey vs. SET filings vs. UN Comtrade — to flag divergences before publication.
Analyst review
A second analyst reviews context, confidence scoring, conflict notes, and limitation handling. Quantitative claims stay tied to specific sources.
Visible limitations
Freshness, coverage gaps, and translation choices are surfaced in-report, not buried in an appendix. Estimates always carry a visible badge.
Canonical inputs
Reports and any scoped exports use the same canonical inputs, so generated files do not drift from the evidence file.
Evidence grades
Every source is graded on a five-tier ladder. The grade is visible on every citation popover, so readers can weigh each claim by the strength of the source behind it.
The issuer produced the data. First choice: Thai government agencies, SET/SEC filings, Royal Gazette rule text, DBD financial statements, BOT licensee registers.
Widely-cited statistical body or international reference (UN Comtrade, FAO FishStat, NOAA, World Bank WDI, IMF Article IV, OECD.Stat). Counts at full weight for market-sizing, demand, and recency gates.
Secondary analytical source — Trade.gov commercial guides, JETRO briefings, commercial aggregators, trade press. Used for corroboration and colour, never as the sole anchor.
Arithmetic combination of named primary drivers (for example, PromptPay value × wallet share). Requires a full derivation record on the claim so readers can inspect each input.
Analyst synthesis when no source qualifies as any of the above. Last resort. Always carries a visible estimate badge and a gap note explaining why.
Seven coverage gates
Before any report publishes, the source pack must clear seven automated gates. The checks run against the source registry; a report that fails any gate is blocked from release until the gap closes.
- 1.
Market sizing
At least two anchor sources on different bases (baht survey vs USD aggregate) triangulate the headline market value.
- 2.
Operator coverage
Every named operator has a primary-grade filing — SET 56-1, SEC 20-F, or DBD annual financial statement for private entities.
- 3.
Rule text
Every regulation cited links to the actual rule text (Royal Gazette, ministry consultation draft, regulator register), not a press summary.
- 4.
Demand + payment
Demand-side anchors sit independently of supply-side market-sizing. Examples by sector: BOT PromptPay + e-payment for ecommerce/fintech; NOAA + EUMOFA for seafood; OIC segment data for insurance; MOTS + TAT arrivals for hospitality; NBTC subscriber data for telecom.
- 5.
Logistics / enabling infra
Every sector has an enabling-infrastructure layer cited as its own gate. Examples: Port Authority of Thailand TEU for trade-exposed sectors, NBTC spectrum auctions for telecom, grid + bancassurance channels for banking/insurance, Thailand-China rail for cross-border logistics, AOT air-cargo for hospitality.
- 6.
Recency
At least two primary sources refreshed inside the last 12 months, with visible freshness grading per source.
- 7.
Artifact depth
Report-tier artifacts require ≥20 source families (target 20–25). Stat artifacts ≥5. Company artifacts ≥3. Prevents structurally-clean-but-shallow research from shipping.
Product outputs
Canonical report inputs feed the web report first. The same normalized data layer can also power scoped buyer exports, public statistics, and later internal reference datasets when those files are generated.
Limitations
What this method does not claim to do. Stated up front so the boundaries are part of the product, not a surprise.
Thailand is the geographic focus. Regional context is included where relevant, but coverage outside Thailand is not comprehensive.
Thai-language sources are usually richer than English coverage, so translation choices matter and are made visible when they matter.
Government series can lag by 6–12 months. Every figure keeps its vintage visible.
Revenue and production estimates for non-listed companies are clearly marked as estimates.
See it in the reports
Browse published reports or the industry map to see how the method translates into finished intelligence.