Hong Kong M&A · Valuation research

DLOM & Control Premium

Discount for Lack of Marketability and Control Premium figures, extracted directly from HKEX Notifiable-Transaction circulars and verified by a local-LLM panel under Claude supervision.

🔬
2026 Q2 preview. 2026 Q2 preview — a standalone slice, isolated from the full multi-year research (not yet released). Period: 1 April – 30 June 2026. Source: HKEX Notifiable-Transaction circulars. All figures are aggregated — this page presents distributions and summary statistics, not individual deal records.
Headline

The quarter at a glance

23.1%
Mean adopted DLOM
median 20.4% · 8.7–42.9%
23%
Mean Control Premium
median 22.4% · 18.1–31.4%
14
Valuations with DLOM
6 with a Control Premium
97
Circulars processed
22 business/equity valuation reports
Spread

Where the figures land

DLOM vs Control Premium — range & mean

Bars span the min–max of adopted figures; the dot marks the mean (μ).

0% 10% 20% 30% 40% 50% DLOM 8.7% 42.9% μ 23.1% Control Premium 18.1% 31.4% μ 23%

DLOM distribution

14 adopted figures, by 5% band

0 1 2 3 4 5 1 5–10 1 10–15 2 15–20 5 20–25 2 25–30 1 30–35 1 35–40 1 40–45

Control Premium distribution

6 adopted figures, by 5% band

0 1 2 2 15–20 2 20–25 1 25–30 1 30–35
Pipeline

From 97 circulars to 14 figures

Extraction funnel

Only business/equity valuation reports carry a marketability discount.

Circulars downloaded 97 Business/equity VR 22 Adopted DLOM/CP 14

Circulars by month

1 April – 30 June 2026

Apr 24 May 30 Jun 43
Methodology mix

How the assets were valued

Valuation approach

Across all valuation subjects in the quarter

Market Approach 18 Asset-Based Approach 15 Income Approach 6 Cost Approach 3 Comparison Approach 1

Valuation type

What was being valued

Property Valuation 26 Equity Share Valuation 22 Business Valuation 14 Plant And Machinery Valuation 3 Intangible Assets Valuation 2 Financial Instrument Valuation 1
How it's made

Method & quality control

Every HKEX Notifiable-Transaction circular for the quarter is downloaded and classified. Documents carrying a business or equity valuation report go through regex extraction (150+ deterministic patterns) and then an LLM comprehension pass. Figures are only adopted when they reconcile to the valuer's concluded value — board-letter and IFA-letter restatements are kept as cross-checks and never double-counted.

  • Local-LLM panel: qwen3.6:35b + gemma4:26b (local) cross-check each other. ≥2-model agreement + verbatim grounding required to adopt a figure.
  • Claude supervisor: Claude (orchestration, adjudication, PDF spot-checks) — reviewed every zero-figure report, adjudicated edge cases against the raw PDFs, and recovered 1 figure the automated panel missed.
  • Rejected: 3 false-positive regex matches were vetoed by the panel and excluded.
  • Coverage: 22 / 22 figure-bearing reports fully processed.

Isolated run This quarter was processed against a separate database, additive downloads, and its own output folder — fully decoupled from the multi-year production dataset, which was left untouched.

Important

Disclaimer

AI-Assisted Research — Disclaimer and Limitation of Liability

AI-assisted research, human sample-checked (not verified in full). This notice is not limited to AI accuracy — it also covers completeness, methodology, interpretation, and scope. AI can misclassify, omit, or misstate figures. Provided "as is," with no warranty and no liability; use at your own risk and verify against the original documents. Findings are provisional and may be amended or withdrawn at any time. Not professional advice.

Read the full disclaimer

Scope of this notice. This disclaimer is not limited to the use of artificial intelligence, nor to the accuracy of individual figures. It applies to the study as a whole — including its completeness, the methodology and interpretation choices made, the scope and period covered, the assumptions relied upon, and the possibility of subsequent change — and to any use of, or reliance on, any part of it.

Preparation. This study was prepared using an automated, AI-assisted process. Figures and related information were extracted from publicly filed documents (HKEX notifiable-transaction circulars) by a software pipeline combining rule-based pattern matching and large language model (LLM) comprehension, with deterministic checks that require each reported figure to be traceable to a verbatim line in its source document. The process was designed, supervised, and refined through substantial human and programming effort.

Human review is on a sample basis only. Human experts reviewed the methodology and checked a sample of the outputs; where a prior human study existed for part of the period, results were compared against it. The dataset as a whole was not, and cannot practically be, verified line-by-line by a human. Absence of a flagged error does not mean a record is correct.

Known limitations of AI. AI and automated systems can and do make mistakes. These include, without limitation: misreading or misclassifying figures (for example, confusing a discount rate, cost of capital, or premium with a different metric), mis-attributing a figure to the wrong party or valuation, omitting figures, misinterpreting context or intent, and generating plausible but incorrect output ("hallucination"). Source documents also vary in quality and formatting, which can affect extraction. Readers are expected to understand the general capabilities and limitations of AI-based extraction before relying on this material.

Limitations beyond AI. Independently of the AI process, the study is subject to ordinary research limitations: it relies on source documents that may themselves contain errors, ambiguities, or omissions; it reflects classification, interpretation, and judgment calls that reasonable professionals might make differently; it is limited to the specific filings, figure types, and period examined and does not purport to be exhaustive; and it reports figures as stated in, or derived from, the source documents without independently verifying the correctness of the underlying valuations.

Right to amend or withdraw. The findings are provisional and may change. Kenneth C.K. Ma reserves the right, at his sole discretion and without notice, to amend, update, correct, restate, or withdraw any part or the whole of this study and its findings at any time. The information is current only as at the stated preparation date and may be superseded without notice.

No warranty. This study is provided on an "as is" and "as available" basis. Kenneth C.K. Ma makes no representation or warranty of any kind, express or implied, as to the accuracy, completeness, reliability, currency, or fitness for any particular purpose of the information contained herein.

No liability; use at your own risk. To the maximum extent permitted by law, Kenneth C.K. Ma accepts no liability for any loss or damage (including direct, indirect, incidental, consequential, or economic loss) arising from any use of, or reliance on, this study or any part of it. Any use is entirely at the reader's own risk.

Not professional advice. Nothing in this study constitutes financial, valuation, investment, legal, accounting, tax, or other professional advice, and it must not be relied upon as a substitute for independent professional judgment. Before relying on any figure, readers should verify it against the original source document and seek appropriate professional advice.

Best efforts. Kenneth C.K. Ma has applied his best endeavours — in both human review and software engineering — to produce accurate and useful results, but this does not amount to any guarantee of accuracy.

Data generated 2026-07-08 from public HKEXnews filings.