The frameworks

Why the multiple dominates

For large, established businesses, earnings are bounded — they can grow 10–30%/yr but rarely double quickly. Multiples are not bounded by reality, they're bounded by sentiment, which can move much faster. The Meta example below makes this concrete: between peak and trough, multiple compression contributed about as much to the price fall as earnings decline, but the multiple was the part that could re-rate fastest when the story changed.

The black-box intuition

A simpler version of the perpetuity formula: imagine a black box that pays $1 every year forever. How much would you pay for it? If your required return is 10%, you'd pay $10 (10×). If your required return is 20%, you'd pay $5 (5×). The multiple is the inverse of your required return. Higher required return → lower multiple you'd accept. The market's multiple is an average of millions of these calculations, blended with growth expectations.

Worked example — same EPS, different multiple

Consider a hypothetical Japanese mid-cap, three-year hold, modelling two scenarios with identical earnings paths.

Scenario A — consensus view, no re-rating. EPS grows 5% per year. Multiple stays flat at 15× (the market's current pricing is correct). Result: price grows in line with earnings.

Year	EPS	Multiple	Price
0	¥100	15×	¥1,500
1	¥105	15×	¥1,575
2	¥110	15×	¥1,650
3	¥116	15×	¥1,740

Total return: +16% over 3 years (~5% annualised). After 3% inflation, barely break-even.

Scenario B — variant perception, multiple re-rates. Same EPS growth (5%/yr) — fundamentals are identical. But the multiple expands from 15× to 22× over three years as the market gradually recognises a moat it had previously discounted. The thesis is on the multiple, not the earnings.

Year	EPS	Multiple	Price
0	¥100	15×	¥1,500
1	¥105	17×	¥1,785
2	¥110	19×	¥2,090
3	¥116	22×	¥2,552

Total return: +70% over 3 years (~19% annualised).

The decomposition. Of the 70% total return, earnings growth contributes 16pp; multiple re-rating contributes the remaining 54pp. The multiple did roughly 3× more work than the earnings. This is what's meant by "multiple dominates" — the same earnings story produced 16% or 70% depending entirely on how the market chose to price each dollar.

The investor's job is to find Scenario B setups: situations where the earnings story is unremarkable but the multiple is depressed relative to the business's actual quality. When the gap closes, the re-rating carries the return. Most alpha-generator trades look like this. The META 2022 trade (covered below) was a sharper version of the same shape — earnings recovered, but the multiple snapped back from 12.5× to 23× and provided most of the recovery.

1. Pricing power → margin durability

Can the business raise prices without losing volume? Ferrari can; commodity producers cannot. Pricing power is the foundational dimension — it's the thing the moats (next section) actually produce.

2. Long-term growth → compounding runway

Is the addressable market large enough that the business can keep growing for years? A great business in a 1× TAM is worth less than a slightly less great business in a 10× TAM — duration matters.

3. Incremental ROIC → reinvestment quality

Each dollar reinvested — what return does it earn? If incremental ROIC is high and the business has runway, growth creates compounding value. If incremental ROIC is below cost of capital, growth destroys value. This is where many "growth" stories rot.

4. Cash conversion → earnings quality

How much of reported earnings becomes actual cash? Persistent gaps between earnings and cash flow are usually a sign of working-capital problems, aggressive revenue recognition, or capex masquerading as opex (or vice versa).

5. Cyclicality → earnings stability

How much do earnings swing with the economy or industry cycle? High cyclicality lowers the multiple the market is willing to pay, because the next reported earnings could be very different from the last.

6. Capital allocation → management quality

What does management do with the cash? Buybacks at low multiples create value; buybacks at peak multiples destroy it. M&A at fair prices in adjacent markets compounds; M&A at premium prices in unrelated markets diworsifies. The capital-allocation track record is one of the cleanest predictors of future returns.

1. Intangible assets

Brand, patents, regulatory licenses, government-granted exclusivity. Ferrari, Hermès, Louis Vuitton, pharma patents, FDA-blessed monopolies, the rights to a piece of intellectual property like Mickey Mouse or Mario. The moat: customers will pay more for the same physical product because of what's stamped on it; competitors literally cannot copy the brand or the patent.

2. Switching costs

Once a customer is committed, the friction of leaving is high. Enterprise software (changing your CRM costs millions in retraining and risk), Bloomberg terminals, banking infrastructure, healthcare-IT systems. The moat: even an inferior competitor's product won't win, because the cost of switching dwarfs the marginal benefit.

3. Network effects

The product becomes more valuable as more people use it. Meta (sharing photos is worth more when everyone's there), Visa/Mastercard (accepted everywhere because everyone uses it because it's accepted everywhere), eBay's core marketplace, language platforms. The moat: a new entrant has to start with zero users and isn't valuable until they have lots, but they can't get them without being valuable.

4. Cost advantage

Structurally cheaper than competitors — through scale, process, location, or proprietary inputs. Costco (membership model + low margins on goods), Walmart (logistics scale), GEICO (direct-to-consumer model), Backblaze (storage cost engineering). The moat: competitors can match the cost only by becoming equally large or restructuring their business model — both very expensive.

5. Efficient scale

The market only supports one or two profitable players. Pipelines (only one needed between point A and point B), regional airports, freight railroads, some utility-style infrastructure. The moat: a second entrant would split the market and make both unprofitable, so no one enters.

The moat-trend diagnostic

Not all moats hold forever. The crucial question is: is this moat widening, stable, or narrowing?

• Widening — the moat is getting stronger over time. (Early-stage Costco, Visa in the 2000s, Meta's network in the 2010s.) These are the best compounders to own; the market typically under-prices duration.
• Stable — the moat holds steady. (Mature consumer brands, established utilities.) Reasonable returns, less excitement.
• Narrowing — the moat is eroding. (Intel against TSMC, legacy retail against Amazon, traditional media against streaming.) Avoid even at apparently cheap multiples — the multiple keeps re-rating down faster than you can collect dividends.

A wide-but-narrowing moat is almost always a worse bet than a narrow-but-widening one. The direction matters more than the current size.

Type 1 — Quality is improving before the market notices

Returns come from both axes: the dot moves right (quality grows) AND up (market re-rates to recognize the new quality). Hold periods are long; the edge is patience plus accuracy on the trajectory. Examples: Costco's first decade as a public company; Meta's recovery 2023–2025; many infrastructure businesses adding new high-ROIC use cases.

Type 2 — Market priced for a decline that doesn't materialize

Quality stays roughly flat; the fear-discount lifts as the predicted decline fails to arrive. Returns are pure re-rating. Hold periods are shorter; the edge is correctly assessing that the bear case is overstated. Examples: TSMC during China-Taiwan fear cycles; O'Reilly during Amazon-fear; the Meta drawdown itself was largely Type 2.

Compounders

Sustained EPS growth over a long runway. You buy a business whose multiple is fair-to-reasonable (you're not expecting it to change), and you ride the earnings growth. Your edge: the market under-prices how long the growth can last. Hold period: years. Temperament required: patience through periods when the business is performing but the multiple isn't re-rating; conviction to hold through volatility that doesn't change the thesis.

Alpha generators

The multiple re-rates, usually with some earnings growth alongside. Your edge: the market is wrong about the company today — excess fear, or quality improving before noticed. Hold period: months to years. Temperament required: contrarianism plus active monitoring; willingness to be wrong publicly during the period before the re-rating happens.

Running both

A balanced portfolio runs both engines. Compounders are the core; alpha generators are the supplement. Compounders give you the "lazy" returns; alpha generators give you the active edge that justifies your fees (literal or opportunity-cost). The split depends on your conviction and your temperament — there's no universal right ratio.

Step 1 — Asset value

What would it cost to rebuild this business from scratch today? Reproduction cost of the physical assets, plus the cost of building the brand, the customer relationships, the distribution. This is the downside floor — if market price is below asset value, you have a free option on everything else the business might be worth.

For asset-light businesses (software, brands) this number is harder to estimate but often much higher than the balance sheet suggests. For asset-heavy businesses (manufacturing, infrastructure) the balance sheet is closer, but needs adjustment for inflation and unrecorded brand/customer value.

Step 2 — Earnings power value (EPV)

What can the current business sustainably earn, capitalized at cost of capital? Strip out one-time items, normalize for the cycle, distinguish maintenance capex from growth capex. EPV = (sustainable earnings) ÷ (cost of capital). If EPV is greater than asset value, the business is earning a return above its capital cost — that excess is the franchise value, and it's only durable if there's a moat.

Step 3 — Growth value

Only counts when management can reinvest at incremental ROIC above cost of capital. If they can't, growth destroys value. If they can, the formula is: Growth Value = (reinvestment rate × ROIC) ÷ (cost of capital × (cost of capital − growth rate)). The intuition: every dollar reinvested at high ROIC creates value above what shareholders could earn elsewhere.

Most valuation mistakes happen here. People assume growth always adds value (it doesn't), or they project growth far beyond what the moat can sustain (it can't). Be conservative; the Step 1 floor is what protects you when you're wrong about Step 3.

Worked example 1 — Hot-dog stand (asset value only)

A single hot-dog stand in a neighbourhood with no franchise. Truck, equipment, signage, inventory: reproduction cost ¥2M. Earns ¥600K/yr after the owner's wage. Cost of capital: 12%.

• Step 1 — Asset value. ¥2M to rebuild from scratch. This is the floor — if the business were trading below this, a buyer could buy and immediately have ¥2M of value.
• Step 2 — EPV. ¥600K ÷ 12% = ¥5M. The current earnings stream, capitalised. EPV (¥5M) is greater than asset value (¥2M), so there's franchise value of ¥3M — but only if the moat is real. For a hot-dog stand with no franchise, no brand, no switching costs, the "moat" is just location and the owner's relationships. Other operators could open a competing stand next door for ¥2M and compete the excess returns away within a few years.
• Step 3 — Growth value. Zero. There's no scalable growth path; one stand is one stand. Even if there were, reinvestment in a competitive market would just be more stands earning the same ~12% — no value added beyond cost of capital.

Honest valuation: closer to asset value (¥2M) than to EPV (¥5M), because the EPV-asset gap will be competed away over time. Maybe ¥2.5–3M reflects the goodwill of an existing operation with established customers.

Worked example 2 — Mid-cap consumer brand (asset + EPV + small growth)

A Japanese mid-cap with a recognisable consumer brand. Replacement cost of physical assets: ¥30B. Brand value (estimated cost to rebuild equivalent brand awareness through marketing): ¥20B. Sustainable earnings: ¥8B/yr. Cost of capital: 8%. Growth runway: ~5% per year over the next 10 years, then fading to GDP+inflation.

• Step 1 — Asset value. ¥30B physical + ¥20B brand reproduction = ¥50B. The floor for a well-known brand is much higher than its book value would suggest.
• Step 2 — EPV. ¥8B ÷ 8% = ¥100B. EPV exceeds asset value by ¥50B — that's the franchise value the moat is generating. For a real brand with pricing power and customer loyalty, this franchise value should be durable.
• Step 3 — Growth value. Assuming reinvestment at 15% incremental ROIC (above the 8% cost of capital) for the next decade, growth value adds roughly ¥30–40B on top.

Honest valuation: ¥130–140B. Asset value is the floor; EPV is the core; growth value is upside that only materialises if management continues to allocate capital well. Buy with a margin of safety against the EPV (e.g., ¥100B or below), treat the growth value as a bonus.

Worked example 3 — High-multiple compounder (growth dominates)

A US software business with a network-effect moat. Asset value: small (capital-light). EPV today: $20B. But the business is growing 25%/yr at incremental ROIC of 40%, with a long runway and limited credible competition.

• Step 1 — Asset value. Maybe $2B for the engineering team, infrastructure, sales relationships. Far below market value — but that's the right answer for an asset-light compounder. Asset value isn't the relevant lens here; it just confirms there's no downside floor to lean on.
• Step 2 — EPV. $20B / cost of capital ~9% = ~$220B. Already a high number relative to current operating performance — the EPV captures the existing earnings stream capitalised.
• Step 3 — Growth value. If the business compounds reinvestment at 40% ROIC for 10 more years and then fades, growth value could be $400–600B on top of EPV. Most of the company's value is in growth value, not in current earnings.

Honest valuation: $600B–$800B if you trust the growth runway and the moat for the next decade. $0–$100B if either the runway or the moat fails. The valuation range is enormous because it depends almost entirely on a long-duration forecast.

This is where most valuation mistakes happen. Investors assume growth always adds value (it doesn't — only growth above cost of capital), and they project growth far beyond what the moat can defend. The discipline: start at Step 1 (asset value floor), work up to Step 2 (current earnings power), and treat Step 3 as the speculative add — not the foundation.

Decomposing the fall

Earnings fell ~46% (peak EPS ~$14 → trough ~$7.5). Price fell ~75%. The multiple compressed from 26× to 12.5× — a -52% re-rating. Roughly half of the drawdown was earnings deterioration; the other half was multiple compression. Most of the recovery had room to be pure re-rating, before any earnings work.

What the market priced in

Four real fears:

• Apple ATT shock — iOS 14.5 broke ad targeting; CFO quantified $10B 2022 revenue headwind. Stock fell 26% next day.
• TikTok / Reels dilution — engagement migrating to short-form video; Reels monetized below Feed.
• Reality Labs burn — $13.7B operating loss in 2022; capex jumped 69% YoY ($18.6B → $31.4B).
• Reputational + macro — Haugen leak; rate hikes compressing all long-duration tech; ad market softening.

Each independently legit. The variant-perception question was whether the combined discount was right-sized, given the fact that several of these were either being addressed (ad targeting) or were optional (Reality Labs).

Was the quality actually declining?

Sum-of-the-parts unbundles it. Family of Apps held 40% segment margins and the network-effect moat — quality intact. Reality Labs was a separable optional bet — could be stopped any time. The market was pricing the consolidated number (FoA minus Reality Labs investment); looked like decline; was actually elective reinvestment on top of a still-very-profitable core.

The personal insight that gave conviction to size the trade: "My friends were on TikTok but sharing the videos back to me on Instagram. They were consuming on TikTok, producing on Instagram. The network-effect moat holds."

What drove the recovery

Four catalysts converged through 2023–2025:

• Year of Efficiency (Q4-22 call) — 21K layoffs, $40B buyback, opex/capex cut. 2023 operating margin +12pp YoY.
• Reels monetization — $1B run rate (Q2-22) → neutral (Q3-23) → $50B annual run rate (Oct 2025).
• AI capex narrative shift — same spending; re-read as moat instead of vanity. AI lifted ad pricing and time-spent.
• Reality Labs tolerated — losses kept growing ($13.7B → $17.7B) but FoA operating income exploded ($42.7B → $89.1B). Market overlooked the burn once the core was visibly working.

Quality didn't decline. The market just learned it had been wrong about the story. This was Type 2 mispricing from earlier on this page — the fear-discount lifted and the underlying kept going.

What generalizes vs. what was luck

Repeatable (the framework): SOTP framing that separated FoA from Reality Labs; recognizing capex as investment not impairment; sizing for re-rating not just earnings recovery; respecting the base rate that founder-led companies tend to defend the core when pressured.

Not repeatable (the luck): exact timing of the Q4-22 efficiency-pivot announcement; the AI tailwind through 2023–24 reframing the capex narrative; speed of sentiment reversal once the catalysts hit.

Conviction is not the same as correctness. The framework is what generalizes; the timing was favorable.