Boom Leverage
Platform due diligence dossier
- Edition
- 04
- Published
- Supersedes
- 03 · 30 September 2026
Companies indexed
926
Scope: Listed Thai companies with at least one customer-reachable Form 56-1 row in a served domain, FY2021+, non-company symbols removed.
Measured under the tightened rule adopted on 2026-09-06; directly comparable to edition 03's 923 (2026-09-30).
[1]Citation gate (5 published axes)
98.86%
Scope: 506,696 of 512,529 findings, Core era FY2021+, across five published axes: auditor, business, financials, governance, risk. Excludes beta, archive and auxiliary domains.
Same five-axis population as edition 03 (489,178 of 494,377). The denominator grew by 18,152 findings between editions.
[2]Findings withheld (5 published axes)
5,833
Scope: Core FY2021+ findings across the five published axes that failed word-for-word PDF verification; complement of the 98.86% pass rate.
Same five-axis population as edition 03's 5,199. A withheld finding is never shown as a verified result.
[2]Markets
1 live · 1 in build
Scope: Thailand serving; Vietnam corpus in build, not yet searchable.
Vietnam pricing is published and the corpus is being built; Vietnam is not yet on sale.
[3]How this document is produced
This dossier is drafted by a language model from a fixed prompt and a supplied evidence pack, then reviewed and published by the founder, who is accountable for every claim in it. It is not an independent third-party opinion and does not pretend to be one. What makes it checkable is the appendix: every figure below carries its scope, the date it was measured, its source, and the way you can re-derive it yourself. Where the evidence pack could not answer a question the structure asks, the question is listed as unanswered rather than filled in.
Questions this edition could not answer
The structure of this dossier asks for the following and the evidence pack did not contain it. They are listed rather than filled in.
- Commercial metrics — revenue, paying accounts by tier and retention — are not disclosed in this edition.
- No third-party security certification (SOC 2 Type II or ISO 27001) is held at this edition.
- Runtime pass rates for the Boom Leverage AI citation validator (chat and /scan summaries) are not yet published; the citation gate figures in this memo measure indexed findings.
- Delivery scope and service levels for API, MCP and Excel add-in integrations are set per design-partner engagement and are not published.
- The launch date for Vietnam search is not set in this edition.
§1
Executive summary
- Boom Leverage indexes the Form 56-1 filings of 926 listed Thai companies [1], serving 729,578 verified-source PDF-quote rows [6]. Across the five published axes (FY2021+), the citation gate verified 98.86% of findings word for word against the source filing — 506,696 of 512,529 — and withheld the remaining 5,833 rather than show them as verified [2]. The FY2016–2020 archive verifies at 99.24% [4] and the beta forensic axis at 98.94% [5].
- Since edition 03 the platform shipped Boom Leverage AI: chat and /scan summaries that answer only from retrieved filing passages, each sentence tied to a cited source and checked by a citation validator before display [7][8].
- Self-serve plans run from FREE (100 credits a day) through DELTA (฿799/month) and GAMMA (฿1,399/month) to ALPHA desk licences from ฿9,900/month [7]. Above them sit On-Demand Research from $1,000 per report and the Thai Stock Case File — Members programme (฿99,000/year, opening January 2027) [9].
- The verification discipline is unchanged: every figure in this memo is dated, scoped and re-checkable by a third party, eras and domains are never pooled into one rate, and a finding that cannot be matched to its page is withheld, not shown.
- Vietnam is the second market: pricing is published [10] and the corpus is in build, with no company searchable yet [3].
§2
What changed since the last edition
Between 2026-09-30 and 2026-11-17 the platform shipped Boom Leverage AI with source-bound answers and a citation validator [7][8], simplified search pricing to 1 credit for one to five axes and 2 credits for six or more [8], and aligned paid Opp Day depth with filing depth — four years on DELTA, six on GAMMA [7]. The storefront gained full Thai pages, enterprise integrations are offered to ALPHA desks as a design partner, and the Thai Stock Case File — Members programme opened for reservations ahead of its January 2027 launch [9]. The Thai corpus moved from 923 to 926 companies [1], with the five-axis gate at 98.86% on a larger denominator [2] across 729,578 served PDF-quote rows [6].
| Item | Previously | Now | Dated |
|---|---|---|---|
| Generative AI search layer (Boom Leverage AI) | No generative synthesis in the search interface | AI chat (/terminal/chat) and /scan AI summaries shipped with post-generation citation validator, one repair pass and weekly output-token limits (30k–1.5M+) | |
| Credit metering schedule | +1 credit per axis beyond 5; AI summary unlaunched | Two-tier axis pricing (1 credit for 1–5 axes, 2 credits flat for 6+); AI summary +2 credits (0 on 24h same-language replay); chat turn equals search cost | |
| Self-serve Opp Day windows | DELTA 8 rounds (~2 years); GAMMA 16 rounds (~4 years) | DELTA 4 years (16 rounds); GAMMA 6 years (24 rounds), aligned with each tier's Form 56-1 depth | |
| Storefront redesign and Thai parity | English-led storefront with partial Thai article translations | Full Thai versions of every page, bilingual guides and 23 Thai research articles | |
| Enterprise integration positioning | API listed on ALPHA desk licence | API, MCP (Claude/ChatGPT) and Excel add-in offered to ALPHA desks as design-partner integrations | |
| Commercial lead offer on /th | Terminal subscription led /th homepage | Pre-launch reservations for Thai Stock Case File — Members (฿99,000/year, January 2027) bundling GAMMA and bespoke 1–4 day deliverables | |
| Core Thai index and citation gate | 923 companies; 489,178 of 494,377 verified (98.95%, 5,199 withheld) across 5 published axes | 926 companies; 506,696 of 512,529 verified (98.86%, 5,833 withheld) across 5 published axes; 729,578 served PDF-quote rows |
§3
The market gap
Thai securities rules require every listed company to file an annual Form 56-1 One Report: hundreds of pages of operating commentary, risk factors, governance disclosure and the auditor's report. Across the 926 companies served [1], the numbers in these filings are already digitised and priced by incumbent terminals; the prose is not read at market scale. Operating deterioration — a shift in customer concentration, a local supply bottleneck, receivables ageing — routinely appears in that prose quarters before it reaches consensus.
The commercial point is that this disclosure is unread, not unavailable. The edge comes from reading mandatory public filings systematically, not from private data. Keyword search fails because management writes around the issue: a search for "customer concentration" misses a filing that says "the majority of the Company's revenue … comes from government and state enterprises customers". Boom Leverage embeds 729,578 served PDF-quote rows [6] with a multilingual model, so a concept query in Thai or English retrieves the passages that describe it in either language. Because the citation gate, not the extractor, decides what is shown, an analyst sees only passages proven against the page — and is told how many were withheld [2].
The value is historical memory, not price forecasting. GAMMA reaches six years back (FY2021+) and ALPHA desk licences reach FY2016–2026 [7], so a desk can find the reporting period in which management's risk framing changed. The FY2016–2020 archive (254,038 verified findings across the five published axes) is measured on its own denominator, so historical backfill never flatters the current pass rate [4][2].
Form 56-1 prose is lawyered; Opportunity Day briefings are not. Spoken guidance at SET Opportunity Day carries context the printed filing leaves out. Boom Leverage indexes the briefings by timestamp so a reader can open the recording at the minute a statement was made, and spoken figures are shown only when the session's own presentation deck corroborates them. Across the rolling 72-month window, 8,044 sessions are in scope and all 7,684 eligible sessions are transcribed and highlighted [12].
§4
Extraction architecture and the bounding box
Every result binds four things: the verbatim passage, the filing it came from, the page number and the position of the passage on that page. Form 56-1 PDFs keep their word coordinates through ingestion and are embedded by a multilingual retrieval model. Where a filing is an image-only scan, OCR and vision models are used only to recover the text and its word positions — never to paraphrase. A filing whose scan cannot yield coordinates, or that has not yet been fetched, produces no verified result.
Before any finding renders, a deterministic citation gate compares the extracted passage character for character with the text at its recorded position. On 2026-11-17, across the five published axes (FY2021+), 506,696 of 512,529 findings passed (98.86%) and 5,833 were withheld [2]. By axis: auditor 99.62% (14,034 of 14,088), risk 99.18% (153,339 of 154,604), governance 99.08% (53,051 of 53,545), business 98.74% (65,380 of 66,213) and financials 98.58% (220,892 of 224,079) [2]. The beta forensic axis passes 98.94% (111,530 of 112,726) [5] and the FY2016–2020 archive 99.24% (254,038 of 255,974) [4], each on its own denominator.
Across all ten PDF-quote domains served — the five published axes plus forensic and four further disclosure domains — 721,047 of 729,578 rows verify (98.83%) [6]. That figure is not comparable with edition 03's 598,776, because the four added domains joined the denominator in this edition [6]. Datasets that are not quoted from a PDF page, such as Opportunity Day recordings and regulatory notices, are linked to their source and reported separately from these rates.
For readers who do not read Thai, every cited row shows the original Thai passage as the authoritative text beside an English translation, with the link to the filing on the SEC's own site. Boom Leverage AI follows the same rule: answers are written only from retrieved passages, every sentence carries its citation, and the validator checks each citation against its source before the answer is shown [8].
§5
The false-positive philosophy
Boom Leverage carries a bank model-risk posture: a false positive costs an institution more credibility than a missed signal. At the index, the citation gate fails closed — when text and position disagree the finding is withheld, not approximated: 5,833 in the five published axes [2], 1,196 in beta forensic [5] and 1,936 in the archive [4]. A withheld finding is never shown as a verified result; a user who wants to inspect near-misses sees them on a separate, labelled shelf behind a warning.
On the auditor axis, extracting every routine liquidity or going-concern note would bury the signal. The parser reads the auditor's wording to separate modified opinions (qualified, adverse, disclaimer) from routine Emphasis of Matter paragraphs. In the 2026-08-15 study across FY2021–FY2025, 68 of 910 companies carried modified opinions, and a deterministic rule moved 96 TSA 570 going-concern template rows out of the red-flag bucket (772 rows to 676), clearing 60 companies without touching any of the 68 modified-opinion flags [11].
Thai speech recognition is weakest on numbers and units. On Opportunity Day transcripts, a spoken figure is shown only when the session's presentation deck corroborates it and its magnitude agrees with the financials corpus, and no severity score is ever assigned to how something was said.
Boom Leverage AI applies the same posture to generated answers. Each answer is restricted to retrieved sources (30 per chat turn; 20, 35 or 45 per /scan summary on FREE, DELTA and GAMMA [7][8]), checked by the citation validator with one repair pass, stripped of any claim it cannot cite, and refunded in credits if it is blocked [8].
§6
Data integrity, isolation and leak prevention
Three leak classes govern data handling: query visibility across clients, generative hallucination and stale figures. Against the first, the public Form 56-1 index is separated from user query histories and from the taxonomies clients apply, and neither is used to train or tune shared models — one desk's searches never shape another desk's results.
Against hallucination, Boom Leverage AI is bounded by scope as well as by citation: it does not calculate, rank or forecast, every summary carries a visible notice that it is AI-generated and should be checked against its sources, and every sentence must survive the citation check before display (up to 400k input tokens per thread and 2,000 output tokens per answer [8]).
Against stale figures, the verification appendix binds every number to its measurement date and population: the 2026-11-17 corpus figures [2] are kept apart from the 2026-08-15 auditor study [11], and a count is never restated as current after the window it measured has changed.
Institutional buyers should note that Boom Leverage does not yet hold a third-party security certification such as SOC 2 Type II or ISO 27001; requirements for ALPHA desk licences (from ฿9,900/month [7]) are scoped with each desk.
§7
Business model and unit economics
Boom Leverage sells an action-metered subscription ladder priced in THB, with annual plans at ten months' price for twelve [7]. FREE gives 100 credits a day over about five quarters of filings and two quarters of Opp Day. DELTA (฿799/month, ฿7,990/year) reaches four years of filings and Opp Day with 150 credits a day and 50-row Excel/CSV export; GAMMA (฿1,399/month) reaches six years with 350 credits a day [7]. ALPHA desk licences start at ฿9,900/month: 1,500 credits a day, history from FY2016 to FY2026, six years of Opp Day, multiple seats and 100-row export per search [7].
Usage is metered by action, so a desk pays for what it reads: a search costs 1 credit for one to five axes and 2 for six or more (+1 per 25 companies beyond 25); an AI summary adds 2 credits, with a free 24-hour replay; opening a boxed PDF quote or an Opp Day clip costs 2; a single-company Health Check costs 3; an auditor-opinion reveal costs 5; and following a citation to the SEC's own site is free [8].
Above the subscriptions sit two research offers. On-Demand Research starts at $1,000 per report, with up to three revisions [9]. The Thai Stock Case File — Members programme (฿99,000/year, opening in the first week of January 2027) bundles twelve chosen stocks (up to 30 swaps a year, delivered one to four days after each filing), twelve eight-year deep-dive requests a year, three private research topics, a document-activity report, 30-day early access and twelve months of GAMMA [9].
Enterprise integrations — API, Claude/ChatGPT through MCP, and an Excel add-in — are offered to ALPHA desks as a design partner: each is built with the desk that will use it. Every desk already receives export files carrying the verbatim source passage and its citation link [7].
§8
Founder and institutional credibility
Boom Leverage is founded and operated by Varanchai Yingkhamnueng, whose relevant pedigree comes from institutional banking model risk rather than commercial software engineering. At Kasikornbank, he worked in credit-risk quantitative analytics and built an NLP early-warning system over Thai news and disclosure prose to detect borrower deterioration ahead of non-performing loans. At Krung Thai Bank, as a Senior Data Scientist in business risk and macro research, he owned and operationalised the bank's monthly TFRS 9 expected-credit-loss dashboard in production and validated it against the consultant's model specification.
That background sets the product's boundaries. Defending a TFRS 9 provisioning process before internal bank audit requires line-item provenance regardless of headline accuracy — which is what the bounding-box citation gate, the TSA 570 boilerplate quarantine [11], the refusal to score tone on Opportunity Day, and the source-bound rules inside Boom Leverage AI [8] deliver here.
The platform is founder-led, and the dependency is real: research direction, the risk taxonomy and client delivery run through one person. It is offset by design rather than by headcount — the citation gate is deterministic, ingestion and verification run as automated pipelines, and every figure a client relies on is re-checkable from the filing without the founder in the loop. Key-person dependency is the first entry in §10.
§9
Expansion thesis
Capability is reported in three states. Shipped: Thai Form 56-1 semantic search and Boom Leverage AI. In progress: API, MCP and Excel add-in integrations, built with ALPHA desks as a design partner. Planned: the Thai Stock Case File — Members programme, opening January 2027 [9]. Regional expansion follows a franchise model: the multilingual retrieval stack and the citation gate are carried to the next market's mandatory filings, starting with Vietnam's annual reports (Báo cáo thường niên) across HOSE, HNX and UPCoM.
In Vietnam the storefront is live and pricing is published at parity with the Thai packages — ₫639,000 a month for DELTA, ₫1,119,000 for GAMMA and from ₫7,900,000 for ALPHA [10] — with a waitlist in place of checkout. The corpus is being built and no Vietnamese company is searchable yet [3]; nothing is sold until it is.
Vietnam opens for sale when three conditions hold: its annual reports yield text and word positions that clear the same citation standard as Thailand's five published axes (98.86% [2]); the multilingual model retrieves Vietnamese disclosure language as reliably as Thai; and Vietnamese desks adopt the published pricing [10] without a separate product.
§10
Risk register and open questions
Five risks follow, each with the mechanism by which it would bite and the evidence that would retire it. They cover key-person dependency, generated answers, source-document quality, security assurance for institutional procurement, and the pace of expansion against incumbent distribution.
01Key-person dependency
- Mechanism
- Research direction, the risk taxonomy and delivery of the research offers run through the founder. A prolonged absence would slow product development and bespoke delivery — most visibly the 1–4 day turnaround promised to Thai Stock Case File members [9] — even though the index, the citation gate and self-serve access continue to run unattended.
- What would answer it
- A named second analyst or engineer in the delivery chain, published runbooks for ingestion and verification, and a stated member capacity for the Case File programme.
02Generated answers between validator and reader
- Mechanism
- A citation validator confirms that each sentence of a Boom Leverage AI answer points at a retrieved source [8]. It cannot fully guarantee that a correctly cited sentence frames its source the way the filing does, so a reader who acts on a summary without opening the citation carries that interpretive risk.
- What would answer it
- Published runtime validator pass and repair rates, with a human-audited accuracy sample of chat and /scan answers.
03Dependence on source-document quality and format
- Mechanism
- Coverage depends on filings yielding text and word positions. A poor scan or a change in the regulator's template can leave a passage unverifiable; the gate withholds it rather than guess — 5,833 findings in the five published axes [2] — so the cost is a missed passage, not a wrong one. An analyst who reads an empty result as proof that a company is silent can still draw the wrong conclusion.
- What would answer it
- Gate rates published separately for native-text and scanned filings, and machine-readable filings adopted by the regulator.
04Security assurance for institutional procurement
- Mechanism
- Boom Leverage separates the public filing index from client query histories and does not train shared models on client queries, but it does not yet hold SOC 2 Type II or ISO 27001. Procurement teams that require a certificate before onboarding may lengthen or block an ALPHA desk licence (from ฿9,900/month [7]).
- What would answer it
- A completed SOC 2 Type II or ISO 27001 audit, or a standard security questionnaire pack accepted by an institutional client.
05Expansion pace against incumbent distribution
- Mechanism
- Vietnam is priced but not yet searchable [3][10], and each new market needs its filings to clear the same citation standard before it is sold. Incumbent data terminals with existing distribution could bundle comparable semantic search over the same public filings before a second market is live.
- What would answer it
- Vietnam live with a multi-year cohort that clears the Thai five-axis standard, and paying desks in a second market.
§11
Verification appendix
Every quantitative claim in this memo links through [[n]] to its value, the exact population counted, its measurement date, its source and a method a third party can use to re-check it. Eras and domains are never pooled into one rate, and a figure measured over a window that has since changed is not restated as current.
| # | Figure | Population counted | Measured | Source | How to re-check it |
|---|---|---|---|---|---|
| 1 | 926 companies (previous: 923 on 2026-09-30) | Listed Thai companies with at least one customer-reachable Form 56-1 row in a served domain, FY2021+, non-company symbols removed. | GET https://engine.boomleverage.com/universes?market=th (count, count_basis) | Query /universes?market=th or execute an unfiltered whole-market scan on the Terminal; compare directly with edition 03's 923 measured under the same 2026-09-06 rule. | |
| 2 | 98.86% citation-gate pass — 506,696 of 512,529 verified, 5,833 withheld (previous: 98.95%, 489,178 of 494,377, 5,199 withheld on 2026-09-30) | Core era FY2021+ served rows across the five published axes: auditor (14,034 of 14,088 = 99.62%), business (65,380 of 66,213 = 98.74%), financials (220,892 of 224,079 = 98.58%), governance (53,051 of 53,545 = 99.08%), risk (153,339 of 154,604 = 99.18%). | terminal-engine/data/index/th/.citation_gate.json (computed_at=2026-11-17T01:56:30Z, index_built_at=2026-11-17T00:29:54.487087+00:00) | Verify 506,696 + 5,833 = 512,529 and sum the five published entries in by_domain inside .citation_gate.json. | |
| 3 | 0 Vietnamese companies indexed | Listed Vietnamese companies across HOSE, HNX and UPCoM currently indexed and searchable. | GET https://engine.boomleverage.com/universes?market=vn → {"error": "not_built"} (no universe.json for market=vn), 2026-11-17 | Search the Terminal with market=vn: no Vietnamese filing is returned; the /vn page sells no plan and runs a waitlist. | |
| 4 | 99.24% archive citation-gate pass — 254,038 of 255,974 verified, 1,936 withheld (plus archive beta and aux blocks) | Archive era FY2016–FY2020 across the five published axes (auditor 12,176/12,200 = 99.80%; business 30,890/30,894 = 99.99%; financials 111,831/112,757 = 99.18%; governance 16,525/16,525 = 100.0%; risk 82,616/83,598 = 98.83%); reported separately from the FY2021+ rate and never added to row 2. | .citation_gate.json, archive block | Inspect the archive object in .citation_gate.json; confirm 254,038 + 1,936 = 255,974 and year_min=2016, year_max=2020. | |
| 5 | 98.94% beta forensic citation gate — 111,530 of 112,726 verified, 1,196 withheld (previous: 99.11%, 103,468 of 104,399 on 2026-09-30) | Beta forensic axis (branded Operating quality on live pages, covering six MD&A lenses), FY2021+; kept on its own denominator outside row 2. | .citation_gate.json, beta block | Inspect beta.forensic in .citation_gate.json; verify 111,530 + 1,196 = 112,726. | |
| 6 | 729,578 served Form 56-1 / PDF-quote rows (721,047 verified, 8,531 withheld, 98.83%; not comparable to edition 03's 598,776) | Served rows FY2021+ across 10 PDF-quote domains (auditor, business, financials, forensic, governance, legal_disputes, related_party, risk, shareholding, strategy) out of 15 domains in _stamp.serve. Edition 03's 598,776 (2026-09-30) measured an earlier domain set without those four auxiliary PDF-quote domains. | GET /universes?market=th (rows_served) and .citation_gate.json (evidence.kinds.pdf-quote, _stamp.serve) | Verify rows_served equals evidence.kinds.pdf-quote.total (729,578) and 721,047 + 8,531 = 729,578. | |
| 7 | FREE 100 credits/day · DELTA ฿799/mo · GAMMA ฿1,399/mo · ALPHA from ฿9,900/mo | Thai Terminal entitlements by tier: FREE = ~5 quarters of filings, 2 quarters of Opp Day, 100 credits/day, 30k AI output tokens/week, no export; DELTA = 4 years, 4 years of Opp Day, 150 credits/day, 500k AI tokens/week, 50-row export; GAMMA = 6 years, 6 years of Opp Day, 350 credits/day, 1.5M AI tokens/week, 50-row export; ALPHA = FY2016–2026, 6 years of Opp Day, 1,500 credits/day, AI limits per agreement, 100-row export. Annual billing = 10 months for 12 (฿7,990/year on DELTA). | boomleverage-web/src/lib/mdna-tiers.ts; boomleverage-terminal/src/lib/ai-quota.ts and mdna-types.ts (export caps); /th/credits | Read MDNA_FREE and MDNA_TIERS in mdna-tiers.ts; compare export limits across /th/product/mdna-search, /th/enterprise and /th/about. | |
| 8 | 1 credit (1–5 axes) · 2 credits flat (6+ axes) · AI summary +2 credits · max 2,000 output tokens/answer | Credit cost per action: search = 1 (1–5 axes) or 2 (6+ axes), +1 per 25 companies beyond 25; AI summary = +2 (free 24-hour replay); chat turn = search price (30 sources per turn); boxed PDF quote = 2; Opp Day clip = 2; single-company Health Check = 3 (+1 per extra year); auditor-opinion reveal = 5; translation = 1; SEC link = 0. AI limits: 2,000 output tokens per answer, 400k input tokens per chat thread. | boomleverage-terminal/src/lib/mdna-credits.ts and ai-quota.ts; /th/credits | Inspect MDNA_CREDIT_COST, MDNA_AXIS_BUNDLE, MDNA_AXIS_OVER_BUNDLE and mdnaAxesCost in mdna-credits.ts; verify discrepancies with /th/credits and /th/product/mdna-search FAQ. | |
| 9 | $1,000/report (On-Demand Research) and ฿99,000/year (Thai Stock Case File — Members, PLANNED Jan 2027) | Published Thai Stock Case File — Members price and offer terms (opening first week of January 2027; reservations open, no payment taken): 12 chosen stocks (up to 30 swaps a year, delivery 1–4 days after filing), 12 eight-year deep-dive requests a year, 3 private topics a year, document-activity report, 30-day early access, 12 months of GAMMA. On-Demand Research: from $1,000 per report, up to 3 revisions. | boomleverage-web/src/lib/research-service.ts; boomleverage-web/src/components/casefile-full-preorder.tsx; /th | Verify $1,000 in research-service.ts and ฿99,000/year as CASEFILE_FULL_PRICE_THB in casefile-full-preorder.tsx; the offer terms are printed on /th. | |
| 10 | ₫639,000/mo DELTA · ₫1,119,000/mo GAMMA · from ₫7,900,000 ALPHA (Vietnam, uncharged) | Published Vietnam tiers, at parity with the Thai packages at 1 THB = 800 VND (annual ₫6,390,000 DELTA and ₫11,190,000 GAMMA, 10 months for 12); waitlist only — no checkout until the corpus is live. | boomleverage-web/src/lib/vn-pricing.ts; /vn | Inspect VN_TIERS and VN_ALPHA_FLOOR in vn-pricing.ts; compare against mdna-tiers.ts scaled by 800. | |
| 11 | 68 of 910 companies (modified opinions) and 96 boilerplate rows quarantined clearing 60 companies (772 → 676 rows, median 240/company) | FROZEN historical audit from the 2026-08-15 auditor-opinion research article covering FY2021–FY2025 modified opinions (qualified, adverse, disclaimer) and TSA 570 going-concern boilerplate quarantine inside the red-flag bucket; not re-measured on the 2026-11-17 corpus of 926 companies. | Edition 03 [9] and [11] (auditor-opinion research article, boomleverage.com) | Verify quoted opinions, regulator links and the before/after table (772 → 676) in the published auditor-opinion article. | |
| 12 | 8,044 Opp Day sessions in the 72-month window; 7,684 eligible, 7,684 transcribed and highlighted (100% of eligible; 95.5% of the window) | SET Opportunity Day sessions with a recording, rolling 72-month window from 2020-11-17. Eligible excludes 360 sessions with no filing to cross-reference (trusts filing a 56-REIT, group seminars, recent listings). Same definition as edition 03's 5,755 sessions, which covered a four-year window on 2026-09-21. | GET https://engine.boomleverage.com/oppday/coverage?market=th (receipt 2026-11-17T00:10:02Z) | The coverage receipt publishes in_window, eligible, done, pending and skip with its window start; the /th page prints the same fill rate live from it. |
Earlier editions
Archived editions are kept verbatim and are not updated. Their figures were correct on the date each was measured and should be quoted with that date attached.
Bespoke Forensic Report
The same discipline, on a name you choose
This dossier is the format applied to ourselves. Commissioned on a company or a question, it is the same memo: findings first, every assertion cited to the filing, page and passage, and the evidence set exported so your analysts can re-run the work instead of trusting it.