The auditor warned for years and nobody read it — 68 Thai listed companies whose audit opinion was not 'clean' (FY2021-FY2025)

Run this scenario.

You pull up a stock. Net income is positive. P/E screens cheap against the peer group. The dividend is still being paid. The note circulating in the group chat says "fundamentals are intact, the price is down on market conditions". So you add to the position.

In the same file as that earnings number — the file you already downloaded — a few pages before the financial statements, a short line sits printed as a heading on the page.

ไม่แสดงความเห็น · Disclaimer of Opinion

The person the law empowers to open the real books, read the real documents, question management directly, and sign their name under legal liability is telling you they cannot form an opinion on this set of numbers. Not that the statements are wrong. That they will not vouch that the statements are right.

And you scrolled past it.

That is not a jab. It happens thousands of times a year, and the reason has nothing to do with how hard you work. We will take it one obstacle at a time.

The brutal part of a cooked-books case is not the day the news breaks

Every time Thai equities hit a big one — statements whose numbers do not match reality, a company whose numbers stay beautiful right up to the day they stop, with no warning — the sequence repeats.

News breaks → floor limit down for days on end → you cannot sell because there is no bid → and afterwards everyone says the same thing: "it was obvious"

That last sentence is the trap. It was obvious because you already know the ending. I wrote this out at length in the early-warning system I built inside a bank, because it is the same error that breaks an entire risk model. The far more expensive question: on that day, what information was actually in your hands?

For a lot of these cases the answer is that the information already existed, was already disclosed, was already free to download. It was not buried in a cryptic hint. It was a heading printed in bold on the page — and it is in both the Thai and the English edition.

So why does retail always take the worst of it? Not because retail is dumber, and not because institutions hold secret data — I will show you the identical information sitting in a public file. It is because the institution has a process that forces someone to open that page. Retail has no such process.

The one page in the book the company did not write

Read this part slowly. It is the entire reason this page carries weight the rest of the book does not.

The whole Form 56-1 (One Report) — vision, strategy, management discussion and analysis (MD&A), risk factors, sustainability — is written by the company itself. It is still worth a great deal; I wrote a whole piece on the four-part, nine-chapter map of the One Report about what you can mine out of it. But never lose sight of what it is: one side's testimony, written under a legal ceiling. Not a verdict.

One section is not that: the report of the certified public auditor, sitting on the first page of Part 3 (the financial statements).

The bottom-right cell is the whole point of this article. If the numbers in the statements cannot be trusted, the P/E you computed off those numbers cannot be trusted either. This is not one more factor to check. It sits above every other factor, because it decides whether the other factors mean anything.

And because it is a fixed template, it is data that genuinely compares across companies and across years — unlike MD&A prose, where every company writes in its own style.

Read it in four lines: the auditor has a finite vocabulary

Four things in the auditor's report matter, and three of the four are not danger signals. Anyone who cannot tell them apart panics at the wrong thing every time. That is why most people quit reading this page altogether.

1. Type of opinion — this is the verdict

A trap that exists only in the English edition: the word Disclaimer, in the language a Thai reader meets every day, means "ข้อความปฏิเสธความรับผิด" [the boilerplate nobody reads] — the thing at the bottom of every email and every website that everyone scrolls past. In this document, a Disclaimer of Opinion is the harshest verdict an auditor can write. The most severe rating in the entire book arrives wearing the costume of the least important sentence in it.

The Thai edition has no such trap — it says "ไม่แสดงความเห็น" [we are not putting our name behind these numbers] flat out, and there is no way to misread it. That is one reason I tell people to read the Thai book (the Thai search strings are in the section "Why retail cannot do this alone" below) · but if you open the English edition — which is the version my Terminal quotes from, and the version cross-border analysts use — this is the one word you never scroll past.

2. Material uncertainty related to going concern

The standard requires the auditor to assess whether the business can keep going for at least 12 more months. If there is significant doubt, it goes in a separate paragraph. It looks like this — actual text from the auditor's report for AAV, FY2021, page 199, during the COVID period when the airline stopped flying:

"This indicates that there are material uncertainties which may raise substantial doubt about the Group's ability to continue as a going concern."

This is the most misread paragraph in the document. It is not a death certificate. AAV still trades today. A going-concern paragraph means there is material uncertainty and the auditor is obliged to tell you. It does not mean this company is going under. Your job when you see it is to read what the uncertainty actually is and whether it was gone the following year. Not to hit sell.

3. Emphasis of Matter paragraph (TSA 706)

The auditor points you at a specific note to the financial statements without changing the opinion. It is a read-this-it-matters marker. In my archive, 496 companies carry at least one across FY2021-FY2025 — more than half the market. Anything more than half the market has is not a red flag. It is a reading assignment.

4. Key Audit Matters (TSA 701)

The most common false alarm. A KAM is the auditor telling you where the audit effort went this year — it is part of a clean report. Transparency, not warning. In my archive KAM is the largest category of everything on the auditor side (7,874 of 14,165 items). Flag KAM red and you flag nearly the entire market red, at which point red means nothing at all.

Two things warrant alarm: a modified opinion and a going-concern paragraph. Everything else is a reading assignment.

The institutional hard rule, and why it is not gatekeeping

I work on the risk side of a bank — the day job is owning and validating credit risk models. I will not claim to know how every Thai fund writes this rule down, because I have not read every firm's investment manual.

What I can say for the side I sit on: in the counterparty risk process I work in, this field gets filled in first, not later. If the numbers were not certified without qualification, running ratios off those same numbers is arithmetic dressed as analysis on a base nobody has confirmed.

The reason this gate comes before the others is not tradition. It is logic:

1. It is the only signal in the document that comes from outside — everything else is the company talking about itself.

2. The writer sees what you will never see — ledgers, invoices, contracts, confirmations from banks and receivables.

3. Lying is expensive for the person signing — their professional licence is on the table.

4. It is comparable data — a fixed standard template, so you can line up the whole market. MD&A prose you cannot.

Point 4 is where retail is furthest behind, and it is the one a tool fixes rather than effort. The arithmetic is below.

So how many companies in the Thai market — I counted

Saying you should read it is cheap. So I counted.

I scanned every auditor's report in the Boom Leverage Terminal index (index built Aug 14, 2026 · counted Aug 15, 2026) — 910 companies with an auditor's report loaded, 14,165 items, covering FY2021-FY2025 (FY2026 is only starting to land). This is what came back:

terminal.boomleverage.com

Auditor's reports, whole market — what the count actually shows

910 companies · 14,165 items · FY2021-FY2025 · counted Aug 15, 2026

• 🚨Red flags, whole market

  187 companies

  20.5% of 910 — the two lines below combined (41 companies overlap)

• 🚨Modified opinions

  68 companies

  ≈ 7.5%, or 1 in 13 · some companies drew more than one type in different years

• ⏳Material uncertainty, going concern

  160 companies

  Real material-uncertainty paragraphs, boilerplate excluded

• ⛔Disclaimer of Opinion

  28 companies

  The auditor gave no opinion on the statements

• ⚠️Qualified Opinion

  45 companies

  Certified except for specific items

• 🔁Flagged 3+ consecutive fiscal years

  17 companies

  Not a one-year accident

• ⚠️Emphasis of Matter

  496 companies

  More than half the market = read it, not a red flag

• 📋Key Audit Matters (KAM)

  7,874 items

  Largest category, and part of a clean report

Two numbers to keep.

One — 187 of 910 companies, 20.5%. That is how many had the auditor write something you should go read: either a modified opinion or a material-uncertainty going-concern paragraph.

Two — 68 of 910, 7.5%, roughly 1 in 13. That is the hard core: the auditor did not certify the statements without qualification. 17 of them were hit for three consecutive fiscal years or more.

Stop there for a second.

7.5% is rare enough to be information. At 60% it would be worth nothing — a burglar alarm that goes off every time someone walks past, which nobody turns to look at. 1 in 13 is the rate at which checking pays, and the check is fast.

And the 17 companies flagged three years running are direct evidence that these warnings do not arrive with the news. They sat on paper for years first.

The staircase down: a case you can read straight off the filings

More powerful than the aggregate is reading one company year by year. The auditor's verdict moves, and the direction it moves is the signal.

Here is ITD (Italian-Thai Development) by fiscal year. Every line is the word printed on the actual auditor's report, with the page number in the filing submitted to the SEC (Thailand's Securities and Exchange Commission):

terminal.boomleverage.com

ITD · Italian-Thai Development

Auditor's opinion type, by fiscal year

• FY FY2021⚠️ Qualified · p.103
• FY FY2022⚠️ Qualified · p.105
• FY FY2023⛔ Disclaimer of Opinion · p.113
• FY FY2024⛔ Disclaimer of Opinion · p.113
• FY FY2025⚠️ Qualified · p.118

“Disclaimer of Opinion — I was engaged to audit the consolidated financial statements of Italian-Thai Development Public Company Limited ("the Company") and its subsidiaries ("the Group") and the separate financial statements of the Company…”

The staircase in one sentence: in FY2021 the auditor had already qualified the opinion. The next year it was still qualified. In FY2023 and FY2024 the auditor stopped giving an opinion at all, before returning to a qualified opinion in FY2025.

The point is not what anyone should do with this stock. The point is timing. The FY2021 document was filed in early 2022. A qualified-opinion-grade signal was sitting in a free public file several quarters before this name became the thing everyone in the market was talking about.

ITD is not alone. Others from the 17 caught in consecutive years — verdict and year only, no commentary from me:

The line, drawn as clearly as I can draw it: a qualified opinion or a disclaimer does not mean the company broke the law, does not mean the books were cooked, and does not mean sell. Plenty of cases are scope limitations — incomplete access to a subsidiary's or an associate's records. A red flag is an instruction to go read the source, not an instruction to pass judgment — and reading the source has to take minutes, not days. That is the problem still left standing.

Why retail cannot do this alone — the arithmetic nobody lays out

By now you may be thinking: just open the file and look at the page. I thought that too, until I did it across the whole market.

Obstacle by obstacle:

1. It is not a language problem — the Thai edition exists, and you should read it. Clearing this first because it is the misconception I meet most: the Thai edition of Form 56-1 (One Report) carries the full auditor's report in Thai in Part 3, and it uses the same fixed standard wording as the English side. Which means Ctrl+F works. These strings are lifted from real Thai filings on my disk. I did not translate them:

Try it on a live one — open IFEC's FY2022 Thai One Report and search ความเห็นอย่างมีเงื่อนไข. Three hits. And in the same book the company describes its own status in a sentence worth reading slowly:

"ตลาดหลักทรัพย์แห่งประเทศไทยได้ประกาศให้บริษัทฯ เข้าสู่ระยะเวลาดำเนินการให้มีคุณสมบัติเพื่อกลับมาซื้อขาย (Resume Stage) กรณีผู้สอบบัญชีรายงานว่าไม่แสดงความเห็นต่องบการเงินเป็นเวลา 3 ปีติดต่อกัน" [three straight years of the auditor refusing to sign, and the exchange has already started the delisting clock on them]

Read it again — "ไม่แสดงความเห็น 3 ปีติดต่อกัน" [the auditor declining to give an opinion three years in a row] is grounds for delisting. That is not my interpretation. It is the company describing itself in a document filed with the SEC, and it answers the question of why institutions treat this as gate one better than anything I could write: the Stock Exchange of Thailand (SET) counts it too.

Another case with the numbers attached: SMK, FY2022. The auditor issued a Disclaimer of Opinion, and the Thai going-concern paragraph (p.81) puts comprehensive loss at ฿32,759 million and states "บริษัทมีหนี้สินรวมสูงกว่าสินทรัพย์รวมจำนวน 30,488 ล้านบาท" [liabilities exceed assets by ฿30,488 million — the equity is gone].

⚠️ The trap that comes with it: search การดำเนินงานต่อเนื่อง on its own and you will hit it in the filings of the cleanest companies in the market, because the auditor's-responsibilities paragraph closing every report carries standard boilerplate on exactly this subject (PCC's Thai book has it, on a clean opinion) — finding the phrase is not finding a problem. You have to check whether it sits in the company's own paragraph or in the template paragraph. This is the same trap that had my own system red-flagging 60 companies incorrectly. Full story below.

So why is the evidence in my Terminal in English? Because the hard rule is: if we cannot draw a box back to the exact source page with 100% certainty, we do not display it. The Thai document set the SEC releases through its financial reporting system usually arrives as .DOC / .XLS files with no page coordinates to anchor to. No provable location = fails the gate = quarantined. So we anchor on the English edition, which arrives as PDF with fixed coordinates, so every line you see clicks back to the original. The quoted text is English because that is what we can prove, not because the Thai does not exist — and when reading on your own, use the search table above against the Thai book. Easier on the eyes.

2. It sits on a different page every year and in every company. Look at the live data above — ITD: pp. 103 / 105 / 113 / 113 / 118 · ANAN: pp. 197 / 201 / 265 · ACAP: p.1, because it filed the statements as a separate document. There is no go-to-page-N rule. You hunt for it every single time, inside a 200-300 page book.

3. Ctrl+F helps less than you think. Search "opinion" in one file and you get dozens of hits, because the word sits in every heading of the report, including the standard paragraphs every company has identically. More on that in the next section, because it is the big trap in any crude count.

4. And here is the multiplier that kills it — you have to read backwards. The lesson from the table above is that direction beats any single year's value. Seeing "qualified this year" on its own tells you far less than knowing whether it just started or has run five years.

Multiply it out:

10 holdings × 5 fiscal years back          =  50 files
~200–300 pages per filing                  ≈  12,500 pages
find the auditor page + read, 5 min/file   ≈  4+ hours
                                              per single sweep

Four hours to reach an answer that 95% of the time is "clean, nothing here" — and you redo it every year, every time you buy a new name, and every time new statements drop.

And here is the part I want unambiguous: the Thai search table above does not reduce that number at all. It helps once the file is already open. The actual work is downloading, opening, hunting for the page, writing down what each year said, and lining them up yourself — 50 times. That is ~25 minutes per company if nothing goes wrong, and it is manual labour, not analysis.

Knowing how to read it and having the hours to read a whole portfolio five years back are two different problems. This article solves the first. The second does not yield to willpower.

This is not a diligence problem. This is textbook information asymmetry. The disclosure is legally equal for everyone. The cost of reaching it is not. The institution has people reading this as a full-time job. You have the free hour at 10pm after work.

Information everyone can reach but only some can afford to collect is still unequal information.

What I built to fix it

Years on the risk side of a bank. The real job is owning the lifecycle of models running in production and validating how far they can be trusted. The single most expensive lesson from that work: an alarm that fires too often is an alarm nobody listens to.

So when that discipline came down into a retail tool, the brief was not "flag everything". It was "flag rarely, flag correctly, and expose every line for audit".

What happened while I was writing this — and why I am telling you instead of staying quiet

While counting the numbers in the table above for this article, I found my own system was red-flagging companies incorrectly.

At the end of the "ความรับผิดชอบของผู้สอบบัญชี" [the responsibilities boilerplate every report closes with] paragraph in every report — including the cleanest in the market — TSA 570 requires this sentence:

"However, future events or conditions may cause the Group and the Company to cease to continue as a going concern."

It is template, not warning. The auditor is not saying this company will not make it. The auditor is saying the standard was followed. My extraction pipeline read it and filed it under going concern — correctly, because that is literally the subject matter — and the red-flag bucket accepted it without asking a second question.

Result: 96 of the 637 rows in that category were pure boilerplate, and they dragged 60 companies with clean signed opinions into the red-flag bucket. Names you know: GULF, INTUCH, BTS, CRC, KCE, SCG and TRUE.

This is the exact failure this whole article warns you about, inverted. Instead of stamping "safe" on a company with a problem, it stamped "dangerous" on companies without one. The end state is identical: the reader stops believing the colour red. And once they stop, the day it turns red for a real reason, nobody looks.

Fixed Aug 15, 2026 with a deterministic rule, not a model confidence score: the sentence is verbatim from the standard, so a rule catches it at 100%. The rule fires only on rows that begin with that sentence, because a genuine material-uncertainty paragraph always opens with the company's own facts first (like AAV above — "…total current liabilities exceeded…") and draws the conclusion afterwards.

The 187 in the table above is the post-fix number. Had I not caught it, you would be reading 246 in this article, then scanning GULF and finding a red flag.

I am telling you because a tool that sells credibility has to show its misses, not only its hits — and because the lesson is the spine of this article: an alarm that fires too often is not safer. It is quietly useless.

The gate I am proudest of: the free tier must never lie that something is "safe"

Slightly technical, but worth telling, because this is where it nearly went wrong.

The free tier sees one fiscal year, the most recent. Now think about ACAP, four straight years of Disclaimer of Opinion. If the card counted only the year a free user can see, and that year happened to carry no item, the card would read "✅ No red flags found" over a company whose auditor refused to sign off four years running.

That is not a missing feature. That is a false safety signal, which is worse than having no tool at all.

So the system is built so counts cross the paywall and content does not — a free user sees "N items exist in years this tier cannot see", which is true and which they need to know. How many red flags exist = free · the evidence (the actual text plus the verified page) = the paid part.

I chose it that way because between hiding the data completely to force an upgrade and telling the truth and letting people decide, the first converts faster. It also sells a false signal, and I will not do that.

What this tool cannot do — read this before you decide

This section matters more than the sales section. If one paragraph survives, make it this one.

1. If the auditor was blindfolded too, the report comes out clean. This is the biggest limitation and I will not talk around it. Fraud sophisticated enough — a full circle of forged documents, counterparties in on it — can clear an audit for years before it surfaces. This tool reads what the auditor wrote. It does not audit the books. Anyone telling you they have a tool that catches every form of cooked books is selling something that does not exist.

2. It is gate one, not the only gate. A clean opinion means passed this gate. It does not mean good stock. There is a great deal left to read.

3. Current data coverage. The archive spans 916 companies in the combined index, of which 910 companies have auditor's reports loaded, across FY2021-FY2026 (FY2025 is the latest complete year). The backfill before FY2021 is still being built and is not searchable yet — that is the real state today, not the state I want.

4. This article is not investment advice. A company named in the tables above means that text exists in a document filed with the SEC. Nothing more than that.

Put it to work in 10 minutes

You do not need my tool. This works on raw PDFs you open yourself. It is only slower — and it works on the Thai edition. The right column is what you paste into Ctrl+F.

(Reading the English edition? Use the equivalents in the table under "Why retail cannot do this alone" above.)

Step 6 is the hardest to do by hand and the most valuable — same reason I laid out in reading PSL's MD&A across years. A single year's value is a dot. The change in wording between years is the signal. Search strings do not help here, because the problem is not finding it. The problem is opening five files and remembering which year said what.

Scan your own portfolio

Doing it by hand: the tables above are complete enough to work without anyone's help. All of it is public, free to download from the SEC, open to everyone.

If you want the answer in 10 seconds instead of 4 hours — Boom Leverage Terminal. Type in the names you hold and look at the 🚨 Red Flag bucket for a number. If one is there, open it for the actual text from the auditor's report, with the page number and a link to the source file at the SEC, and check it with your own eyes.

Start with the position you are most confident about. That is the one you stopped checking the longest ago.

Log in today and the DELTA tier is free through Sept 30, 2026 — a full 4 years of history (deep enough to see a staircase like the ITD table above) · 150 credits/day · no card required. Billing is switched off entirely for the duration, and when the promotion ends the account reverts to the free tier on its own with nothing charged — why I am giving it away, and the full terms · for what the paid tiers actually differ on, read the straight guide to picking a tier · team or institutional (seats · Excel export · API), talk to us at the Enterprise page or contact@boomleverage.com

More on what else Form 56-1 gives up — the full-book map, four parts and nine chapters · why meaning-based search is not Ctrl+F — how semantic search differs from exact-match · why one signal is never enough, and how many lights have to be on before it is genuinely abnormal — one red flag was never enough

Disclaimer: This article is produced for education and to explain how to read documents that listed companies disclose publicly. It is not investment advice, it does not recommend any security, and it guarantees no outcome. All company information cited is lifted from auditor's reports in documents filed with the SEC, with the fiscal year and page number given so you can verify it · a qualified opinion or a disclaimer of opinion is not an indication that a company broke the law or dressed its books · all statistics are measured from an index built Aug 14, 2026 and counted Aug 15, 2026, and will change as new documents enter the archive · investment carries risk, and past results do not guarantee the future.