The Arche Framework
Arche is the firm’s proprietary framework for reading a business. This essay describes how it works.
Company pages provide reported financial facts — revenue flows, margins, returns on capital, growth and ownership. This essay explains how the framework would combine those facts with estimates and judgment.
The framework runs in two deliberately separate steps: first judge the quality of the business, then examine what expectations are embedded in its price.
- The Arche Quality Grade (A–F) scores the business itself across four pillars. Price never enters this grade.
- “What the price implies” (Discounted / Fair / Full / Demanding) reports how today’s market price compares to a deliberately conservative estimate of value — separately, so a great-but-expensive business is never confused with a mediocre one, and the site never publishes anything resembling a ranked buy list.
Gate 0 — The Circle of Competence (“Too Hard” flag)
Arche begins with knowability. A sound analysis must explain how a company makes money, why customers choose it, what capital it requires, and what could break its economics. The size of an analyst’s circle is less important than honestly marking its boundary.
Some businesses are flagged Too Hard before scoring: fewer than ten years of financial history; cumulative owner earnings negative over a decade; economics dominated by a single binary event (a drug approval, a commodity price); or a business model that resists plain-English explanation. The flag is educational, never a condemnation. It marks where this framework’s confidence honestly ends: understandability, durable economics, capable stewardship, and a sensible price must all survive before a conclusion deserves confidence.
Pillar 1 — Business Quality (30%)
| Criterion | Measured as | Strong / Acceptable | Arche rationale |
|---|---|---|---|
| Return on invested capital | 10-yr median ROIC | ≥15% / ≥10% | Durable advantages should protect attractive returns on the capital required to operate. |
| Return on equity | 10-yr median ROE | ≥15% / ≥10% | Returns on owners’ capital reveal more than EPS growth that can arise automatically from retained earnings. |
| Owner-earnings quality | 10-yr owner earnings ÷ net income | ≥90% / ≥70% | Reported earnings matter only to the extent they convert into cash an owner could remove without weakening the business. |
| Margin stability (pricing power) | 10-yr gross-margin coefficient of variation | ≤0.08 / ≤0.15 | Stable margins are evidence that customers value the offering and the business can absorb cost pressure. |
| Earnings consistency | Down years in owner earnings (10 yrs) | ≤1 / ≤3 | Arche favors demonstrated earning power over turnarounds and projections. |
| Capital intensity | 10-yr capex ÷ operating cash flow | ≤30% / ≤50% | The strongest businesses can grow earnings without continuously rebuilding their asset base. |
Owner earnings here = net income + depreciation & amortization − stock compensation − maintenance capex (proxied as the lower of current capex and its five-year average). Stock compensation is treated as a real expense because issuing shares transfers part of each continuing owner’s claim.
Pillar 2 — Moat Evidence (20%)
Half quantitative fingerprint, half written assessment (the written half can never move a grade more than one step on its own):
- ROIC persistence — years of the last ten with ROIC above 10% (10/10 strong). An advantage that must be rebuilt constantly is not durable.
- Revenue consistency — R² of log-linear 10-yr revenue growth (≥0.90 strong): franchises grow steadily, commodity businesses lurch.
- Cash conversion — 10-yr free cash flow ÷ net income (≥90% strong): real advantages should produce cash, not only accrual earnings.
- (Report-judged) moat classification — brand/intangible, switching costs, network effects, low-cost scale, efficient scale, regulatory license, or none — with verbatim evidence quoted from the company’s own 10-K, plus the inversion question: what would kill this business?
Pillar 3 — Financial Fortress (15%)
| Criterion | Measured as | Strong / Acceptable | Arche rationale |
|---|---|---|---|
| Interest coverage | 3-yr avg EBIT ÷ interest expense | ≥10× / ≥5× | Leverage can turn a long record of good operating results into a permanent loss, so debt service needs a wide cushion. |
| Debt payoff power | Total debt ÷ 3-yr avg net income | ≤2 yrs / ≤4 yrs | A resilient business should be able to retire its debt from a few years of normalized earnings. |
| Stock-compensation discipline | 3-yr SBC ÷ operating cash flow | ≤5% / ≤10% | Stock compensation is compensation; the resulting dilution is an economic cost to continuing owners. |
EBITDA appears nowhere in this framework because depreciation reflects the consumption of real assets. Where leverage needs measuring, Arche uses interest coverage against operating income; where cash generation needs measuring, it uses owner earnings after maintenance capital requirements.
Pillar 4 — Management & Capital Allocation (20%)
| Criterion | Measured as | Strong / Acceptable | Arche rationale |
|---|---|---|---|
| The $1 test | Market-value added ÷ retained earnings over the window | ≥1.5 / ≥1.0 | Every dollar retained should create at least one dollar of market value over a sufficiently long measurement window. |
| Share-count trajectory | 10-yr share CAGR | ≤0% / ≤1% | Buybacks below conservative value raise each continuing owner’s stake; dilution does the reverse. |
| Dividend record | Consecutive maintained-or-raised years | ≥10 / ≥5 — or full retention justified by ROIC ≥15% | A sound distribution policy is consistent, rational, and aligned with the returns available inside the business. |
| (Report-judged) buyback price discipline | Repurchases correlated with the stock’s own valuation percentile | — | Repurchases create value only when made below a conservative estimate of business value. |
| (Report-judged) candor | MD&A plain-language and failure-owning rubric | — | Trustworthy reporting explains results plainly, owns mistakes, and applies the same standards in good years and bad. |
Pillar 5 — What the Price Implies (reported separately)
Two conservative, fully disclosed calculations:
- Owner-earnings DCF: base = 3-yr average owner earnings; stage-1 growth = historical growth × 0.75, capped at 10%; 10 years; terminal growth 2.5%; discount rate a flat 10% hurdle. Financial companies use an excess-return model on book value; REITs an FFO proxy.
- Reverse DCF: the growth rate today’s price implies under the same assumptions, shown against the growth the last decade actually delivered.
Bands: Discounted (price ≥20% below the conservative estimate), Fair, Full, Demanding (≥20% above). Business quality and price discipline are both necessary, but they answer different questions and remain separate. These are illustrative estimates under stated assumptions — never price targets.
Sector rulebooks
Standard manufacturing/service metrics mislead for leveraged financial models, so variants apply automatically:
- Banks & insurers — ROA (≥1.2% strong), ROE, efficiency ratio, equity/assets cushion, earnings smoothness; excess-return valuation on book value. Gross margin, ROIC, FCF and EBITDA-style leverage are suppressed as not meaningful.
- REITs — FFO-style owner earnings (depreciation added back), payout judged against FFO, the $1 test waived (distribution is mandatory).
- Utilities — return expectations tempered to regulated reality; leverage norms relaxed.
- Commodity producers — full-cycle averages only; many land honestly in the Too-Hard pile because a commodity price can dominate the economics.
Grading
Pillar scores renormalize over available checks (a data gap is shown as a data gap, never silently counted as a failure). Composite → A ≥85% · A− ≥78 · B ≥70 · C ≥55 · D ≥40 · F below. If more than 40% of a company’s checks lack data, the card is flagged insufficient data instead of graded.
Honest limitations
Quantitative history is a rear-view mirror; moats erode, managers change, and ten years of numbers cannot capture a pending disruption. The framework also inherits its vendor data’s imperfections (documented per metric with quality flags). That is why every grade links to its raw checks — and why the first question, do you understand the business?, belongs to the reader, not to any scoring engine.