Moat / Industry-Structure Analyst

The diagnosed hole: the v2 conviction rubric uses moat as a necessary gate (a conviction-ceiling input, not the final verdict — see "MOAT DOES NOT SET ITS OWN PRICE" below) but the dossier step does not force a structured taxonomy + durability verdict. Analysts were writing "has a moat" without distinguishing Tren Griffin's five sources, Dorsey's structural durability, or Mauboussin's contestability test. This produced INDIAMART overstated as FOREVER-quality when its network effect is contestable (Justdial proved the moat does not exclude), and IGI being soft-called narrow when its trust franchise is the actual moat. This skill makes the moat call explicit, taxonomised, and durability-tested so conviction is grounded, not inferred.

STEP 0 — Pull the canon (binding-safe)

The moat-method atomics are in three canon folders. Pull them at the conviction step:

set -a && source /Users/pw/invest/.env && set +a && \
/Users/pw/invest/.venv/bin/python /Users/pw/invest/data/scripts/32_consult_brain.py \
  --company "<name + moat terms, e.g. 'INDIAMART network effect switching costs contestable'>" \
  --model <model> --step conviction \
  --json-out extracted/grilling/<TICKER>_moat.json

Use --corpus blended so the consult routes to BOTH the academic canon (vault/canon/tren-griffin-moats/, vault/canon/pat-dorsey-moats/, vault/canon/mauboussin-moat/) AND the binding Munger/Buffett moat cluster, surfacing the desk-synthesis atomics ([[scale-economics-shared-as-self-reinforcing-moat]], [[narrative-moat-fails-the-financial-signature-test]], [[deposit-franchise-is-a-banks-cost-of-funds-moat]], [[moat-width-does-not-set-its-own-price]]) alongside them. The blend keeps the CIO primary and down-weights the pitch-side perspective voices (0.85). Cite only slugs actually returned in principles[].slug. If principles is thin → flag it and treat the moat as unconfirmed (cap conviction at 2, no FOREVER).

Erosion is a binding-CIO judgment, not a canon footnote. The durability section below MUST cite at least one of the six binding erosion atoms — [[moats-are-hard-to-maintain]], [[recognize-the-limits-of-moats]], [[fast-moats-can-be-lost-fast]], [[moats-are-constantly-under-threat]], [[avoid-commoditized-businesses]] — and, where erosion is found, the CIO's own sell-discipline action atom [[sell-when-competitive-advantage-erodes]]. This wires the erosion finding directly to what the binding arbiter would do about it.

Also run the binding consult at --step inversion for any FOREVER candidate — pull disconfirming moat-erosion principles before endorsing the moat as lifelong.

India retrieval blindness (known gap — seed around it). The India perspective moat atoms are currently NOT tagged analyst_role: moat, so a --role moat filter never surfaces them — Terry Smith's scale-matters-in-durability / dominant-position-pricing-power / valuation-discipline-relative-to-quality, SOIC's premiumization-financialization-as-moat, and Saurabh Mukherjea's consistent-compounder lens stay invisible. Two mitigations: (1) a one-time, narrowly scoped frontmatter retag pass to add analyst_role: moat to those specific atoms (review each — do NOT bulk-sed, or non-moat atoms get mis-tagged and pollute the role filter for the capital-allocation / behavioral desks); and (2) until that lands, add those terms as explicit --company seeds so the cold consult still pulls them. These voices enter strictly as CONTEXT/lenses (down-weighted 0.85) — they never upgrade a moat tier on their own; the binding arbiter ([[do-not-be-charmed-by-moats-alone]], [[moats-are-hard-to-maintain]], [[avoid-commoditized-businesses]]) sets the verdict.

TAXONOMY — seven sources (apply all that fire)

Source the taxonomy from the canon. Map each claimed moat to exactly one (or two) of:

Dorsey fourth-source check (four-moat-sources-categorization): for each source confirmed above, test whether it is structural (embedded in the business model) or merely transient (market position, first-mover, or size). Only structural sources earn moat credit.

Mauboussin sources-of-added-value (sources-of-added-value-categorization): confirm the moat maps to a genuine source of added value in the industry structure — innovation lead, supply-side advantage, demand-side advantage, or regulatory barrier. A moat with no industry-structure anchor is brand storytelling, not an economic moat.

Bank / lender sub-clause ([[deposit-franchise-is-a-banks-cost-of-funds-moat]]; binding atom [[underappreciated-moats-in-commodity-businesses]]): for a bank or NBFC, the moat is NOT captured by the generic pricing-power test — it is the low-cost sticky deposit franchise (CASA / cost of funds). A lender that funds itself cheaper than rivals and retains those deposits through cycles has the durable advantage; one chasing bulk/term deposits at market rates does not. Identify the franchise here, then route the actual metric (CASA %, cost of funds vs peers, deposit stickiness, Ke) to the bank-valuation sibling rather than running the four-question pricing-power test.

PRICING-POWER TEST (the single sharpest moat signal)

A business with a real moat can raise prices without losing material volume.

Run the four-question test (from pricing-power-test, pricing-power-assessment):

1. Has the company raised prices above input-cost inflation over a 5-year window?
2. Did volumes hold or grow through the price rise?
3. Would customers bear meaningful pain (time, money, risk) to switch to a cheaper alternative?
4. Do gross margins trend stable or expanding over 5+ years (not just 1-yr blip)?

All four YES → strong pricing power, moat confirmed. Three YES → moderate pricing power, narrow moat. Two or fewer YES → pricing power is weak or unproven, moat is contestable or none.

The Tren Griffin pricing-power cross-check (pricing-power-test-8): "Would I rather own the pricing power or the product?" If the answer is the pricing power, the moat is real. If the answer is the product (commodity/feature), the moat is missing regardless of market share.

DURABILITY TEST — widening vs eroding

Moat age + trend + contestability determines whether a stock earns FOREVER or COMPOUND.

Step 1 — Empirical backtest (moat-durability-empirical-backtest, moat-durability-analysis):

• ROIC vs WACC spread: Is ROIC persistently above cost of capital over 5–10 years? Persistence =

the moat is real. Mean-reversion to cost of capital = the moat is being competed away (regression-to-the-mean-test, quantitative-moat-test-roic-wacc-spread).

• The financial-signature test for a NARRATIVE moat

([[narrative-moat-fails-the-financial-signature-test]]; [[owner-earnings-vs-accounting-earnings]]): a real moat leaves a signature in owner-earnings-based returns on capital, not just in the story. A moat asserted in the narrative (brand, "ecosystem", network) but with no persistent ROIC-above-WACC signature is treated as UNPROVEN — cap conviction, do not award it "Narrow" as a consolation tier. Use owner-earnings (not GAAP) in the numerator so accrual-manufactured "quality" (see forensic Dimension 6) cannot fake the signature.

• Market share stability: Has the company's share been stable or rising over 5 years?

(market-share-stability-metric). Falling share under a claimed moat = moat is eroding.

Step 2 — Five-forces structural check (five-forces-industry-attractiveness):

• Supplier power / buyer power / threat of substitutes / new entrant barriers / rivalry intensity.

A durable moat requires at least two forces working for the firm. Pure market-share leadership without structural protection is a moat on paper, not in cash flows.

Step 3 — Barriers-to-entry check (barriers-to-entry-analysis):

• List the specific barriers: capital requirements, regulatory licences, brand incumbency,

switching-cost lock-in, network density. For each, ask: "Could a well-funded new entrant replicate this in 5 years?" If YES → the barrier is low, moat is contestable.

Step 4 — Disruption-risk flag (disruption-risk-evaluation):

• Is a lower-cost, lower-feature disruptor attacking the bottom of the market?
• Is technology compressing the cost of replication? (E.g., AI compressing human-expertise moats.)
• A moat that cannot survive a technology shift is durable in the short term only.

Step 5 — Inversion (moat-durability-inversion-check, management-vs-moat-inversion-test):

• State the strongest case that the moat is not durable: the one failure mode that invalidates

the thesis. If that failure mode is plausible within 10 years, the moat earns NARROW not WIDE, and COMPOUND not FOREVER.

DURABILITY VERDICT (output of the test)

THE INDIAMART PROBLEM (the contestable-moat archetype)

INDIAMART's B2B marketplace has a two-sided network effect that is real but weak: buyers and sellers both show up, but the value (supplier discovery) is searchable — Justdial, TradeIndia, and Google Maps erode it without needing to match scale. The switching cost on the buyer side is near zero (a buyer can search two platforms in 30 seconds). The seller-side lock-in is moderate (paid subscription), not high. Classification: Network effect + weak switching cost, Contestable, Stable-to-Eroding. Correct verdict: COMPOUND-not-FOREVER, conviction capped at 3.

THE IGI PROBLEM (the trust-fragile archetype)

IGI's gemological certification carries a 50-year brand and lab accreditation that is genuinely hard to replicate — the moat source is Intangibles (trust + regulatory recognition). But trust moats are fragile: one high-profile false certification scandal destroys decades of brand equity in months. The moat is Wide today, but the fragility risk makes it Stable-not-Widening. Classification: Intangibles (trust/brand), Wide, Stable. Correct verdict: COMPOUND or FOREVER eligible, but the inversion (fraud/scandal) must be explicitly stress-tested before FOREVER is granted; conviction 4 unless inversion is dismissed.

MAPPING TO THE CONVICTION RUBRIC

This skill enriches the existing STAGE 3 conviction grading. The moat verdict feeds conviction as:

The conviction rubric's hard gates still apply on top: weak FCF → cap 2; ROIC < ~12% → cap 2; bank below Ke → cap 2. The moat tier is necessary but not sufficient for high conviction.

MOAT DOES NOT SET ITS OWN PRICE (ceiling input, not the verdict)

The moat tier is a conviction-ceiling input, not the final gate. It remains NECESSARY (None → cap 1) but is never SUFFICIENT — the valuation + temperament arbiter binds on top. Hold the binding contradiction in view: [[pay-up-for-extraordinary-businesses]] (a wide+widening moat justifies a higher multiple) sits against [[do-not-be-charmed-by-moats-alone]] (the price still binds). A Wide+Widening moat raises the conviction ceiling; it does not set the verdict — the valuation-dcf-longrunway model plus the binding temperament arbiter retain the final call. A great moat at a bad price is still capped: "pay up" means accept a fuller multiple for genuine quality, never "pay any price." Smith-style quality-at-a-price reasoning is paired with the CIO's price-still-binds discipline, not used to justify paying up indefinitely.

OUTPUT (feeds STAGE 3 conviction and STAGE 4 inversion gate)

Return a structured moat block embedded in the conviction/dossier output:

moat_source: [list of active sources from the seven-source taxonomy]
moat_tier: Wide | Narrow | Contestable | None
durability: Widening | Stable | Eroding
pricing_power: Strong | Moderate | Weak
contestability_note: <one sentence on what a well-funded rival would need to replicate>
inversion_risk: <the single most plausible moat-destruction scenario>
conviction_ceiling: <max conviction this moat justifies, before FCF/ROIC gates>
forever_eligible: yes | no  (no if Contestable or Eroding or inversion_risk is plausible <10yr)
cites_principles: [only slugs returned by 32_consult_brain.py — never fabricated]

Hard rules

1. Never call a moat "wide" without passing the pricing-power test (all four questions). Narrative

descriptions of market leadership are not moat evidence.

1. Contestable moats (replicated by a well-funded rival in 5 years) cap conviction at 3 and

permanently block FOREVER — do not override this with qualitative enthusiasm.

1. The ROIC-vs-WACC spread is the quantitative moat test (quantitative-moat-test-roic-wacc-spread).

A business with ROIC persistently at or below WACC has no moat, regardless of what management says.

1. Always run the inversion step for any Wide moat claim: state the one scenario that destroys the

moat. If that scenario is plausible within 10 years, downgrade to Narrow or Stable.

1. Cite only slugs the canon consult actually returns. If the moat canon is thin for this stock →

flag it, treat as unconfirmed, and cap conviction at 2.

1. The Munger/Buffett binding CIO remains the final arbiter of moat/verdict. This skill supplies

the taxonomy and durability test; the CIO judges whether the moat is real enough to act on.