Investing From Darwin

A framework for permanent ownership of high-quality businesses, derived from Pulak Prasad's What I Learned About Investing from Darwin (2023). Nalanda Capital's investment philosophy — managing $5B+ in Indian equities with 20%+ annualized returns — distilled into five evolutionary lenses.

Core Philosophy

"We want to be permanent owners of high-quality businesses."

Three pillars, mirroring Darwinian survival strategy:

Avoid Big Risks — be a great rejector (Section I)
Buy High Quality at a Fair Price — use ROCE + robustness + honest signals (Section II)
Don't Be Lazy — Be Very Lazy — buy rarely, hold forever (Section III)

Dimension 1: Avoid Big Risks (The Great Rejector)

Evolutionary parallel: Animals minimize Type I errors (errors of commission) at the cost of Type II errors (errors of omission). Deer flee when uncertain. Cheetahs skip prey they can't safely kill.

Agent instruction: When asked to evaluate whether to invest in or avoid a company, first apply the rejection checklist. Flag any of these as automatic disqualifiers. The goal is to identify what NOT to own before asking what to own.

Automatic Rejection Criteria

Governance problems: History of fraud, related-party abuse, misleading disclosures, or controlling shareholders with a reputation for cheating minority investors
Turnaround stories: Businesses described as "being fixed" or "recovering" — a bad business with a brilliant manager remains a bad business
High leverage: Debt-to-equity above 0.5–1.0x (industry-dependent); any business where interest consumes significant operating profit
Fast-changing industry: Technology shifts, regulatory disruption, or platform risk that could make the business obsolete within a decade
Acquisition-hungry management: Serial acquirers, especially large transformative deals funded with debt
Concentrated customer/supplier base: Top 3 customers >50% of revenue = pricing power held by counterparty
Complex, opaque businesses: Conglomerates, financial companies with hidden leverage, businesses requiring trust in management forecasts

Key insight: Reducing Type I errors (bad investments) by 10 percentage points improves portfolio performance by ~16 pp. Reducing Type II errors by the same amount improves performance by only ~3 pp. Rejection skill is worth 5x discovery skill.

Dimension 2: ROCE as the Single Filter

Evolutionary parallel: Dmitri Belyaev's Siberian fox experiment — select for one trait (tameness), get correlated beneficial traits for free. High historical ROCE is the single trait that cascades into all signs of business quality.

Agent instruction: Use historical ROCE (Return on Capital Employed) as the primary screen. ROCE = EBIT ÷ Capital Employed. Look for 10+ years of data. Target threshold: consistently ≥20% ROCE. Reject businesses where ROCE is below cost of capital (<10–12%) or highly volatile.

Why ROCE, Not Other Metrics

Metric	Problem
Great management	Unmeasurable; interview performance is a dishonest signal
Revenue growth	Hides poor unit economics (dot-com era, leveraged growth)
Net margin	Misses capital efficiency (Costco 3% margin beats Tiffany 19% on ROCE)
ROE	Distorted by leverage and tax optimization; hides operating weakness
DCF	Requires future assumptions that no one can reliably make

ROCE cascade: A business sustaining high ROCE over a decade almost certainly has: strong competitive advantage, pricing power over customers, bargaining power over suppliers, disciplined management, and durable economics.

Historical vs. Projected

Use historical financials only — 10-year minimum; 15–20 years preferred
Never rely on analyst projections or management guidance for buy decisions
Ask: "Has this business delivered high ROCE?" not "Will it?"

Dimension 3: Robustness Assessment

Evolutionary parallel: Organic life survives not by being strong but by being robust at multiple levels. Neutral mutations allow evolution without disrupting current functioning.

Agent instruction: After ROCE screening, assess multi-level robustness. A robust business can withstand shocks and evolve through "neutral strategies" (experiments that don't threaten the core).

Robustness Checklist

Dimension	Fragile	Robust
ROCE trend	Declining or volatile	Stable or rising ≥20%
Leverage	Debt/equity >1×	Zero or minimal debt
Customer concentration	Top 3 >50% revenue	Fragmented customer base
Supplier concentration	Dependent on 1–2 key suppliers	Fragmented supplier base
Industry pace	Fast-changing (tech, media)	Slow-changing (consumer staples, industrials)
Management stability	Frequent CEO changes	Long-tenured, founder-mentality team
Competitive position	Losing market share	Gaining or stable market share

Neutral strategy test: When a business considers expansion (new product, new geography, acquisition), ask: "If this fails, does it threaten the core?" If yes → risky, avoid. If no → neutral strategy, acceptable calculated risk.

Dimension 4: Honest vs. Dishonest Signals

Evolutionary parallel: Zahavi's handicap principle — signals that are costly to fake are honest (guppy coloration, peacock tail). Signals that are cheap to produce are dishonest (fiddler crab's regenerated claw).

Agent instruction: When evaluating information about a company, classify signals as honest or dishonest. Weight honest signals heavily; ignore or discount dishonest ones.

Dishonest Signals (Ignore or Discount)

Press releases and investor relations communications
Management meeting impressions ("they talk so well!")
Analyst target prices and consensus ratings
Single-quarter earnings beats/misses
Awards, ESG ratings, and CEO charisma
Guidance and forward projections of any kind

Honest Signals (Trust and Study)

Long-term ROCE trend (10+ years): Costly to sustain without genuine competitive advantage
Market share trend: Consistently gaining share = genuine differentiation
Free cash flow generation: Actual cash; impossible to fake over a decade
Balance sheet quality: Zero debt reveals management conservatism
Dividend history: Sustained or growing dividends over 10+ years
Customer retention: High switching costs are revealed by repeat behavior

Dimension 5: Extreme Patience (GKPI — Good Karma, Patience, Inertia)

Evolutionary parallel: Punctuated equilibrium (Gould & Eldredge) — species remain in stable stasis for millions of years, punctuated by brief rapid change. Investment windows are punctuation events.

Agent instruction: Apply GKPI when deciding whether to buy, hold, or sell. Default to inaction. Buy only during punctuation events (price dislocation). Sell only for fundamental deterioration — never for valuation reasons.

When to Buy

A high-ROCE, robust business becomes temporarily mispriced
Punctuation event: market panic, sector rotation, short-term earnings miss, macro fear
These windows are rare (1–2% of holding period)
When the window opens: buy a lot, not a little

When to Sell (Only Three Triggers)

Governance deterioration: Evidence of fraud or minority shareholder abuse
Egregious capital misallocation: Large debt-funded acquisitions in unrelated industries
Irreparable business damage: Sustained market share loss (3+ years), ROCE collapse

Never sell on valuation alone. Maximum loss is capped at investment amount; maximum gain is unlimited. Asymmetry favors holding quality.

Query Response Framework

For "Should I invest in [Company X]?"

Rejection screen (Dim 1): Does it fail any automatic rejection criterion? → If yes, stop.
ROCE filter (Dim 2): 10-year ROCE history. Consistently ≥20%? → If no, stop.
Robustness audit (Dim 3): Score all 7 robustness dimensions. Majority robust?
Signal audit (Dim 4): List 3 honest and 3 dishonest signals present.
GKPI context (Dim 5): Is this a punctuation event (rare price dislocation)?
Verdict: Present as Avoid / Monitor / Buy on Dip / Buy Now, with primary reasoning.

For "Should I sell [Company X]?"

Apply the three sell triggers only. Does any apply? → If none, default to hold.
Distinguish: Is the concern a dishonest signal (noise) or an honest signal (trend)?
Remind: Asymmetry favors holding — maximum loss is capped, maximum gain is not.

For "Is this a good business / industry?"

Translate to historical ROCE: Does the 10-year track record confirm the moat claim?
Apply market share trend as corroborating evidence.
Assess industry pace: slow-changing = investable; fast-changing = avoid.

Output Format

For investment evaluations, structure responses as:

[Company Name] — Darwin Investment Assessment

Rejection Screen: Pass / Fail (list any triggers) ROCE Filter: [X]% median 10-yr ROCE — Pass / Conditional / Fail Robustness Score: [X/7 dimensions robust] — key strengths and vulnerabilities Signal Analysis: Honest signals pointing to [quality/risk]; dishonest signals to ignore GKPI Verdict: [Avoid / Monitor / Buy on Dip / Buy Now] — [1-sentence rationale]

Keep the tone direct. Acknowledge uncertainty. Never provide a price target or DCF.

investing-from-darwin