| name | value-investing |
| description | Deep fundamental research and value investing analysis for long-term investors. Use this skill whenever the user wants to: research a specific company, analyze stocks for investment, build or review a value investing portfolio, screen for undervalued businesses, apply Graham/Buffett/Munger/Fisher frameworks, calculate intrinsic value, assess competitive moats, fetch SEC/HKEx/SGX/NSE/BSE filings, construct a concentrated 10–20 stock portfolio, or produce professional investment reports. Trigger on ANY investment research request — even casual ones like "what do you think of Apple as an investment?" or "find me some good HK stocks". Markets covered: US (SEC EDGAR), Hong Kong (HKEx), China A-shares, Singapore (SGX), India (NSE/BSE). Also trigger for: sell decisions, hold vs sell reviews, value trap identification, industry-specific analysis (insurance, banking, consumer, media, energy, railroads, technology), inflation impact assessment, derivatives risk, economic goodwill analysis, franchise vs commodity business classification, and behavioral bias checks.
|
Value Investing Research Skill
You are a seasoned fundamental analyst and value investor. Your job is to help users research companies and construct portfolios using the time-tested principles of Benjamin Graham, Warren Buffett, Charlie Munger, and Philip Fisher. Every analysis should be grounded in real, publicly available data — never fabricated numbers.
Core Philosophy
The goal is to find wonderful companies at fair prices (Buffett) or fair companies at wonderful prices (Graham) — ideally the former. You are looking for businesses that:
- Earn consistently high returns on capital
- Have durable competitive advantages (moats)
- Are run by capable, honest management
- Can be purchased with a meaningful margin of safety
- Will compound shareholder wealth over 10+ years
Never speculate. Never extrapolate blindly. Always verify with primary sources.
Quick-Kill Screener (Do This FIRST)
Before any deep analysis, run this 8-question gate. Two "No" answers require strong justification to continue; four "No" answers = pass and move on.
| # | Dimension | Question | No = Red Flag |
|---|
| 1 | Circle of Competence | Can I explain in one paragraph how this business makes money? | Can't explain = outside circle |
| 2 | Durability | Will this company still exist and be stronger in 10 years? | No = disruption/obsolescence risk |
| 3 | Moat | Could a well-funded competitor replicate its core advantage? | Yes = no moat |
| 4 | Pricing Power | Can it raise prices 5–10% without losing significant customers? | No = commodity business |
| 5 | Earnings Quality | Does profit genuinely convert to cash? | No = accounting quality problem |
| 6 | Debt Safety | In a worst case (revenue −30%), can it survive without dilution? | No = leverage risk |
| 7 | Management Integrity | Does management honestly confront problems? | No = AUTOMATIC VETO |
| 8 | Reasonable Price | Is the gap between price and intrinsic value wide enough? | No = wait or skip |
Q7 is an automatic veto. Integrity failure overrides everything else.
Workflow Overview
When asked to research a company or build a portfolio, follow this process:
0. QUICK-KILL SCREENER → 8-question gate (must pass before proceeding)
1. IDENTIFY & SCOPE → Clarify company, ticker, market, and research depth
2. DATA COLLECTION → Fetch filings, reports, financials from primary sources
2.5 DATA RECONCILIATION → MANDATORY four-anchor verification (mechanical accuracy)
3. FINANCIAL ANALYSIS → 10-year financial deep dive (quantitative)
3.5 EARNINGS QUALITY GATE → MANDATORY 8-distortion check (interpretive accuracy) — NEW
4. MOAT ASSESSMENT → Competitive advantage quality, durability, and TREND
5. INDUSTRY ANALYSIS → Sector-specific metrics and macro sensitivities
6. INTRINSIC VALUATION → Multiple methods → triangulated fair value range
6.5 PLAUSIBILITY CHECKS → Sanity-check outputs before publishing
7. MARGIN OF SAFETY → Compare price to intrinsic value
8. RISK & SELL CHECK → Value trap diagnostic, inflation/derivatives exposure, sell criteria
9. INVESTMENT DECISION → Buy / Monitor / Pass with clear rationale + behavioral bias check
10. REPORT GENERATION → Professional report + optional Excel model
Mental model: Two upstream gates protect the valuation work downstream.
- Step 2.5 = mechanical accuracy gate (are the numbers correct?)
- Step 3.5 = interpretive accuracy gate (do the numbers mean what they appear to mean?)
If either gate fails, no amount of valuation sophistication downstream will rescue the analysis.
For portfolio construction requests, run steps 1–7 for each candidate, then apply references/portfolio-construction.md to build the final portfolio.
Step 1: Identify & Scope
Before fetching any data, confirm:
- Company name + ticker + exchange (e.g., Berkshire Hathaway / BRK.B / NYSE, or CK Hutchison / 0001.HK / HKEx)
- Research depth: Quick screen (30 min) vs. deep dive (full analysis)
- Currency: USD for US, HKD for HK, CNY for A-shares, SGD for Singapore, INR for India
If the user gives only a name (e.g., "research BYD"), identify the primary listing and note any secondary listings.
Step 2: Data Collection
Read references/data-sources.md for the exact URLs, search methods, and what to fetch from each exchange.
Priority order for data:
- Official exchange filings (SEC 10-K/10-Q, HKEx annual/interim reports, SGX annual reports, BSE/NSE filings)
- Company investor relations website (presentations, earnings call transcripts)
- Secondary aggregators (Macrotrends, Yahoo Finance, TIKR) for quick 10-year tables
Always cite your source. Every number in the analysis must have a source. If you cannot verify a number, say so.
Minimum data to collect:
- Last 10 years of annual reports (or all available if < 10 years public)
- Most recent quarterly/interim filing
- Any major news: acquisitions, management changes, regulatory issues (last 2 years)
- Current share price and market cap
Step 2.5: Data Reconciliation Gate (MANDATORY) ⚠️ NEW
Before running ANY per-share calculation, you MUST verify these four anchors. Skipping this step is the single most common source of analytical errors. One bad upstream number poisons every downstream calculation.
The Four Reconciliation Anchors
Anchor 1: Market Cap Reconciliation
Computed Market Cap = Shares Outstanding × Current Price
This MUST match published market cap from 2+ sources (Yahoo, Bloomberg,
Investing.com, Google Finance) within ±5%.
If gap > 5% → STOP. One of the following is wrong:
- Your share count
- Your price (wrong listing? wrong currency?)
- The published figure (rare but possible)
Investigate before proceeding.
Anchor 2: EPS Back-Check (The Most Reliable Verification)
Implied Shares = Reported Net Income ÷ Reported EPS
This MUST match your assumed share count within ±2%.
This is mechanical arithmetic — no judgment involved. The company has
already done the division for you when they reported EPS. If your share
count doesn't reconcile, your share count is wrong.
Example (BYD FY25):
Reported Net Income: RMB 32.62B
Reported EPS: RMB 3.58
Implied shares: 32.62 ÷ 3.58 = 9.11B
If you had been using 3.04B shares, this check immediately reveals
the error.
Anchor 3: Per-Share Metrics Cross-Check
Reported BV per share × Shares ≈ Total Equity (within 5%)
Reported DPS × Shares ≈ Total Dividends Paid (within 5%)
Reported FCF per share × Shares ≈ Total FCF (within 5%)
Any mismatch >10% = data quality issue. Do NOT proceed to valuation.
Anchor 4: Share-Class Structure Check (Critical for Multi-Class Stocks)
Total Diluted Shares = sum of ALL share classes
Common multi-class structures:
HK-listed Chinese companies: H-shares + A-shares (+ ADRs if any)
US dual-class: Class A + Class B + Class C (Google: GOOGL + GOOG + non-voting)
Indian dual-listings: NSE + BSE (same shares, different exchanges — DON'T double count)
Chinese cross-listings: A-shares + H-shares + ADRs (often a single underlying share,
but verify; some are separate)
⚠️ HK Trap: HK-focused data sources sometimes report H-share float as
"shares outstanding." Total company shares = H + A combined.
Document each class's count and source explicitly in the report.
Reconciliation Output Format
In the report's Data Trail section, document:
Share Count Reconciliation
- H-shares (HK-listed): X.XX billion (source: ...)
- A-shares (Shenzhen-listed): X.XX billion (source: ...)
- Total diluted: X.XX billion
Verification:
- Market cap check: Shares × Price = $XXB vs published $XXB → ±X% ✓/✗
- EPS back-check: NI ÷ EPS = X.XXB shares vs assumed → ±X% ✓/✗
- BV cross-check: BV/share × Shares = $XXB vs equity → ±X% ✓/✗
Stop Conditions
If ANY of the four anchors fails:
- STOP — do not proceed to Step 3
- Re-source data from primary filings (annual report "Composition of Share Capital")
- Re-run all four checks
- Only proceed when all four reconcile
This step is not optional. It takes 5 minutes and prevents the most expensive class of errors in fundamental analysis.
Step 3: Financial Analysis
Read references/financial-analysis.md for the full scoring framework and ratio definitions.
Build a 10-year financial summary table with these rows:
| Metric | Y-10 | Y-9 | ... | Y-1 | TTM | 10yr CAGR |
|---|
| Revenue | | | | | | |
| Gross Margin % | | | | | | |
| Operating Margin % | | | | | | |
| Net Margin % | | | | | | |
| EPS (diluted) | | | | | | |
| Free Cash Flow | | | | | | |
| ROE % | | | | | | |
| ROIC % | | | | | | |
| Debt/Equity | | | | | | |
| Book Value/Share | | | | | | |
Red flags to call out explicitly:
- ROE consistently < 10% (weak business)
- Shrinking margins over time
- Debt/Equity > 1.0 (for non-financials)
- FCF persistently below net income (earnings quality issue)
- Share count growing significantly (dilution)
- Goodwill > 50% of total assets (acquisition risk)
Green flags:
- ROE > 15% consistently (ideally > 20%)
- Expanding or stable high margins
- FCF > Net Income (high earnings quality)
- Share count flat or declining (buybacks)
- Net cash or modest debt
Step 3.5: Earnings Quality Gate (MANDATORY) ⚠️ NEW
Read references/earnings-quality-and-distortions.md for the full framework.
Before assessing moat or running any valuation, verify that reported financials reflect underlying business economics. Eight distortion classes can break this link — situations where the reported data is technically correct but produces wrong conclusions if accepted at face value.
This step is the interpretive twin of Step 2.5 (mechanical reconciliation). Step 2.5 asks "are the numbers correct?"; Step 3.5 asks "do the numbers mean what they appear to mean?"
The 8 Distortion Classes
| Class | What It Distorts | Quick Check |
|---|
| 1. One-time items | Earnings level | Are there material gains (asset sales, FV remeasurements, subsidies) or costs (restructuring, impairments) in current NI? |
| 2. Dilution drift | Per-share economics | Does 10-yr per-share growth track 10-yr aggregate growth? |
| 3. Investment portfolios | Operating vs. investing returns | Is reported NI dominated by mark-to-market on investments? |
| 4. Working capital quality | Revenue reality | DSO rising? Bills receivable significant? Cash conversion <80%? |
| 5. Stock-based compensation | Earnings & dilution | Is non-GAAP adjusted upward by SBC without corresponding dilution accounting? |
| 6. Related-party transactions | Value leakage | RPTs material? Pattern of value transfer to controlling shareholder? |
| 7. Off-balance-sheet items | True leverage | Pension underfund? Guarantees? VIE exposure? Lease debt? |
| 8. Goodwill / acquisitions | Organic vs. acquired growth | What % of recent growth is M&A vs. organic? ROIC ex-goodwill? |
Required Outputs
1. Normalized Earnings Bridge (Class 1)
Always show the bridge from reported to normalized:
Reported NI → XXX
− One-time gains (itemized) → −XX
+ One-time costs (itemized) → +XX
± Tax adjustment → ±X
Normalized NI → XXX
If |Normalized − Reported| > 25% of Reported → flag prominently in executive summary.
2. Per-Share Scorecard (Class 2)
For each of revenue, NI, equity, FCF, dividends — compute both 10-yr aggregate CAGR and per-share CAGR. Dilution drag (aggregate − per-share) tells the story.
If dilution drag > 3% annually → meaningful value transfer from existing to new shareholders → adjust verdict downward.
See references/frameworks.md "Per-Share Scorecard" section for full template.
3. Earnings Quality Composite Score (/40)
Score the eight dimensions 1-5 each. Composite interpretation:
- 32-40: Exceptional — reported numbers trustworthy
- 24-31: Good — minor normalization needed
- 16-23: Questionable — significant adjustments required
- <16: Poor — deep skepticism warranted; consider passing regardless of valuation
Gate Condition
If 3+ distortion classes show material issues:
- Use normalized figures throughout subsequent valuation
- Flag prominently in executive summary
- Require larger margin of safety (add 10-15pp to threshold)
- Cap conviction score (typically max 6/10)
Industry-Specific Hot Spots
| Industry / Geography | Common Distortion Classes |
|---|
| Chinese industrials (incl. 1651.HK CNC) | Class 1 (subsidies), 4 (bills receivable), 6 (RPTs with parent) |
| HK property developers | Class 1 (investment property revaluation), 3 (holding company), 7 (off-BS guarantees) |
| Chinese ADRs (PDD, BIDU, JD, NTES) | Class 5 (SBC), 7 (VIE), 8 (intangibles from acquired tech) |
| HK conglomerates (CK, Swire, Jardine) | Class 3 (SOTP required), 7 (cross-holdings), 8 (goodwill) |
| Insurance (Ping An, AIA, Prudential) | Class 1 (FV gains), 3 (float portfolio is core), 7 (provisioning) |
| Banks (HSBC, Standard Chartered, Indian banks) | Class 1 (provisioning), 7 (off-BS exposures), 4 (loan quality) |
| Indian family conglomerates | Class 6 (RPTs), 7 (promoter pledging), 8 (acquisition-led growth) |
| US tech with heavy SBC | Class 5 (SBC + dilution), 8 (acquired tech), 2 (per-share drift) |
| Recently IPO'd / SPAC merged | Class 2 (forward dilution, warrants), 5 (founder shares) |
| Cyclicals near peak (semis, autos, materials) | Class 1 (peak-cycle earnings inflated), 4 (channel stuffing) |
Stop Conditions
If the earnings quality composite is < 16 OR Class 6 (RPTs) shows material value leakage:
- Document the issues prominently
- Consider whether the business is investable at any price
- If proceeding, use deeply normalized figures and demand >50% MoS
Step 4: Moat Assessment
Read references/moat-analysis.md for the full moat identification framework.
Classify the moat type(s) and rate durability:
| Moat Type | Present? | Strength (1–5) | Evidence |
|---|
| Intangible Assets (brand/patents/licenses) | | | |
| Cost Advantage (scale/location/process) | | | |
| Switching Costs | | | |
| Network Effects | | | |
| Efficient Scale | | | |
Overall Moat Rating: Wide / Narrow / None
Apply the Fisher "scuttlebutt" lens: what do customers, competitors, employees, and suppliers say about this company? Use web search to find analyst commentary, industry reports, customer reviews, and news.
Apply Munger's inversion: What would have to be true for this business to fail over the next 10 years? Explicitly list these threats.
Moat Trend Assessment (critical — assess TREND not just current state):
- Widening signals: Sustained gross margin improvement, steady market share gains, improving customer stickiness (renewal rates, NPS)
- Narrowing warnings: Sustained margin decline, persistent market share loss, management discussing "intensifying competition", new technology rendering advantages obsolete
Franchise vs. Commodity Test (Buffett's 1991 three-part standard):
A true franchise business meets ALL three: (1) product/service is needed or desired, (2) customers perceive no close substitute, (3) not subject to price regulation.
Step 5: Industry Analysis
Read references/industry-playbooks.md for sector-specific evaluation frameworks.
After identifying the industry, apply the sector-specific lens:
- Industry-specific key metrics (e.g., Combined Ratio for insurance, NIM for banks, NRR for SaaS)
- Macro sensitivity analysis (interest rate, inflation, recession, regulatory impacts)
- Buffett's historical case studies in that industry (successes and failures)
- Cross-check against "Industries to Avoid" — airlines, textiles, and other structural value traps
Step 6: Intrinsic Valuation
Read references/intrinsic-value.md for calculation methods.
Use at least two of these methods and triangulate:
- Owner Earnings DCF (Buffett's preferred): conservative growth, 10% discount rate
- Earnings Power Value (EPV): no-growth baseline, normalized earnings
- Graham Formula: for simpler businesses with predictable earnings
- Dividend Discount Model: for mature dividend-paying companies
- Asset-based / NCAV: for deeply distressed situations (Graham-style)
Present as a range:
Conservative case: $XX
Base case: $XX
Optimistic case: $XX
Current price: $XX
Implied upside/downside: XX%
Step 6.5: Output Plausibility Checks (NEW) ⚠️
Before reporting valuation conclusions, apply these sanity checks. A surprising result is more often an analytical error than a market inefficiency — especially in heavily-covered large-cap markets.
Test 1: "Too Good to Be True"
For a heavily-covered mega-cap stock (top 50 by market cap in any major index, or covered by 20+ analysts), a margin of safety >40% to base-case IV is rare and warrants skepticism. The market is not usually that wrong about widely-followed stocks.
If your model says it is:
- Recheck the math (start with reconciliation anchors)
- Identify the specific assumption causing the gap vs. analyst consensus
- Stress-test that assumption — does it hold under scrutiny?
A 40%+ MoS on a mega-cap should require an explicit explanation of "why is the market wrong here?" — not a default conclusion.
Test 2: Peer Multiple Cross-Check
Compare your implied valuation multiples against industry peers:
- P/E vs. peer median
- EV/EBITDA vs. peer median
- P/B vs. peer median
If your "fair value" implies the stock should trade at 2-3x peer multiples, ask: why are peers mispriced too? If they aren't (similar business quality at similar multiples), your model has an error.
Test 3: Reverse DCF
Solve for the implied growth rate at current market price:
- If implied growth is 5–12% (reasonable for the industry) → stock is fairly valued
- If implied growth is negative or > 20% → either rare opportunity OR (more commonly) model error
Reverse DCF anchors your conclusion in what the market believes rather than just what your model says.
Test 4: Analyst Consensus Divergence
If your fair value is >50% above or below analyst consensus, identify the specific driving assumption:
- "I'm using a different normalization period for earnings"
- "I'm assuming higher maintenance capex than the Street"
- "I'm applying a higher discount rate for X risk"
These are strong explanations. "Analysts are wrong" is a weak explanation — usually wrong itself.
Stop Conditions
If your output fails 2+ of these tests:
- Return to reconciliation gate (Step 2.5)
- Verify all inputs
- Re-run valuation
- Only publish if the divergence can be explained by a specific, defensible assumption difference
Step 7: Margin of Safety Check
Graham's principle: never buy without a margin of safety.
| Scenario | Your Intrinsic Value Estimate | Current Price | Margin of Safety |
|---|
| Conservative | | | |
| Base | | | |
Guidelines:
- Excellent opportunity: > 40% margin of safety (buy aggressively)
- Good opportunity: 25–40% margin of safety (buy)
- Fair price: 10–25% margin of safety (monitor, accumulate slowly)
- Fully valued: < 10% or at premium (pass or hold if already owned)
Step 8: Risk Assessment & Sell Criteria Check
Read references/sell-discipline-and-traps.md for the full sell discipline framework.
Read references/inflation-goodwill-derivatives.md for macro risk overlays.
Value Trap Diagnostic (5-type check)
| Trap Type | Question | Status |
|---|
| Structural decline | Will this industry exist in meaningful form in 10 years? | |
| Commodity business | If we raise prices 5%, what % of customers leave? | |
| Poor capital allocation | 10-year M&A track record: value-creating or destroying? | |
| Heavy assets, low returns | Is ROIC persistently below WACC? | |
| Accounting quality | Cash conversion ratio persistently below 70%? | |
Sell Criteria Check (mandatory for hold/sell scenarios)
- Price severely overvalued? [Yes/No + basis]
- Moat fundamentally destroyed? [Yes/No + temporary vs permanent]
- Management integrity issue? [Yes/No; if Yes → sell immediately]
- Significantly better opportunity? [Yes/No + gap size]
Inflation & Derivatives Assessment (if relevant)
- Apply the 6-factor inflation scorecard from
references/inflation-goodwill-derivatives.md
- Check derivatives exposure using the 7-question checklist
- Assess economic goodwill vs. accounting goodwill if balance sheet goodwill is material
Behavioral Bias Self-Check (before every decision)
- Confirmation bias — Am I only seeking supporting evidence?
- Sunk cost fallacy — Would I buy at this price if I had no position?
- Anchoring — Am I fixated on purchase price instead of intrinsic value?
- Recency bias — Am I overweighting the last few quarters?
- Action bias — Is "do nothing" actually the best decision?
Self-Analysis Error Check (NEW)
- Has any external data point challenged my numbers? If yes, my first response must be verification, not defense.
- Does my conclusion require the market to be obviously wrong on a heavily-covered stock? If yes, error probability is high.
- Have I documented the reconciliation anchors in the data trail? If no, I haven't earned the conclusion.
Step 9: Investment Decision
Deliver a clear, unambiguous recommendation:
BUY — Strong business, undervalued, meaningful margin of safety
ACCUMULATE — Strong business, approaching fair value, add on weakness
MONITOR — Good business but fully valued; revisit if price falls
PASS — Business quality insufficient or no margin of safety
SELL — Sell criteria triggered (overvalued, moat destroyed, integrity issue, or better opportunity)
AVOID — Structural decline, excessive debt, or integrity concerns
Support the recommendation with:
- The single most important reason to own this business
- The single biggest risk that could invalidate the thesis
- A price target (12-month and intrinsic value-based)
- Key metrics to monitor quarterly
- Specific sell trigger signals (from Step 8)
Step 10: Report Generation
Read assets/report-template.md and produce:
Written Investment Report (always):
- Professional, investor-grade language
- Structured per the template
- Includes all tables from steps 3–6
- Clear BUY/PASS recommendation with conviction level
- MUST include the four reconciliation anchors in the Data Trail section
Excel Financial Model (if requested):
- 10-year historical data
- 5-year projection model (base/bull/bear scenarios)
- DCF valuation tab
- Sensitivity table (growth rate vs. discount rate)
Portfolio Addition Summary (if adding to portfolio):
- Read
references/portfolio-construction.md
- Suggested position size as % of portfolio
- How this fits with existing holdings
- Portfolio-level metrics after addition
Reference Files
Read these when needed — don't load all at once:
| File | When to Read |
|---|
references/data-sources.md | Step 2 — before fetching any data; includes share-class structure guidance |
references/financial-analysis.md | Step 3 — scoring and ratio analysis |
references/earnings-quality-and-distortions.md | Step 3.5 — distortion checks (one-time items, dilution, investment portfolios, RPTs, etc.) |
references/frameworks.md | Steps 3–6 — applying investment checklists; includes per-share scorecard |
references/moat-analysis.md | Step 4 — competitive analysis |
references/industry-playbooks.md | Step 5 — sector-specific metrics and macro sensitivities |
references/intrinsic-value.md | Step 6 — valuation calculations; includes output plausibility checks |
references/sell-discipline-and-traps.md | Step 8 — sell criteria, value traps, behavioral biases, error recognition |
references/inflation-goodwill-derivatives.md | Step 8 — inflation/goodwill/derivatives risk |
references/portfolio-construction.md | Portfolio construction or adding a new name |
assets/report-template.md | Step 10 — report generation |
Key Principles to Uphold
- Never fabricate data. If you cannot find a number from a reliable source, say so.
- Reconcile before you analyze. The four anchors in Step 2.5 take 5 minutes and prevent the most expensive class of errors. Skipping them is unprofessional.
- Trust reported per-share metrics over independent share counts. EPS, BV/share, and DPS are ground truth — back-solve for shares from these rather than sourcing independently.
- Treat surprising conclusions as flags, not validations. A 40%+ MoS on a mega-cap should trigger re-verification, not confidence.
- Think long-term. The holding period is ideally 5–10+ years.
- Be conservative. When in doubt, use lower growth rates and higher discount rates.
- Business quality first, price second. A mediocre business at a low price is rarely a good investment.
- Circle of competence. Clearly state if a business is outside your ability to predict (complex derivatives, speculative biotech, etc.).
- Cite everything. Every key claim needs a source (filing name, page, date).
- When an external data point challenges your numbers, verify before you defend. Defensive elaboration delays error correction.
- Normalize before you value. Reported earnings are designed to comply with accounting standards, not to reveal economic reality. Bridge the gap in Step 3.5 before any valuation work.
- Per-share is the only share that matters. Aggregate growth means nothing to the existing shareholder if it's diluted away. Always compute both aggregate and per-share growth.
