Three Floor-Price Methods — Why They Diverge & How to Choose

The /data/floor-price page returns three numbers labelled Dividend Floor, Valuation Floor, and EPV Floor. They almost always disagree — sometimes by 30%+ — and the EPV value usually prints lowest. That isn't a bug; it's the point. Each floor answers a different question, and reading them as a triangulation rather than a single answer is what makes the page useful.

The three questions, three answers

Why EPV almost always prints lowest

Three structural reasons stack on top of each other:

1. No-growth perpetuity. EPV uses NOPAT / WACC. A normal DCF would use NOPAT / (WACC − g) with terminal growth around 2–3%. Removing that g raises the denominator from ~5–6% to 8%, cutting the value 33–67% before you do anything else.
2. 5-year averaged EBIT, not forward. Dividend Floor and Valuation Floor look forward (next year's dividend, next year's EPS). EPV looks back five years and averages. A company in recovery mode prints lower under EPV than under the other two.
3. Punitive 8% WACC. The default cost-of-capital is intentionally conservative for a hard floor. Each 1pp added to WACC lops ~12–15% off the value under perpetuity math.

How to read the three together

Don't pick one. Read the spread.

1. Tight cluster (within ~10%)

Common for stable dividend blue chips — KO, PG, ABT, ENB. Take the median for your Sell Put strike. The methods agree because the company's growth rate is small and predictable, so the no-growth assumption costs you very little.

2. Wide spread, EPV much lower (30%+ gap)

The market is pricing in real growth that hasn't yet shown up in 5-year average EBIT. Examples: a software company past inflection, a cyclical in early recovery, a consumer brand expanding internationally. The truth is somewhere between Valuation Floor and EPV. If you're a buyer demanding margin of safety, weight EPV; if you're writing puts and just need a sane strike, weight Valuation Floor.

3. EPV is null, the other two diverge wildly

Either the company isn't consistently profitable on a 5-year basis, or NOPAT − debt is negative. Don't use this page as your sole anchor — these are the names where DCF, multiples, and option premiums all disagree because the future cash flows are not yet observable. Lower confidence by default.

4. Dividend Floor is null but the others compute

Non-dividend payer, or fewer than 5 consecutive years of dividends. Lean on Valuation Floor for the optimistic anchor and EPV for the conservative one; the median of two is a fine working strike.

What this page is, and what it isn't

It is a strike anchor for selling puts and a conviction check for adding to a position on a drawdown. It is not an intrinsic value model — none of the three floors estimate "fair" price. They estimate floors: prices at which historical evidence (Dividend), market behaviour (Valuation), or arithmetic (EPV) say a buyer should be very interested.

A stock trading 5% above its three floors clustered tightly is uninteresting from a Sell Put standpoint. A stock trading near the median of its three floors with EPV providing real downside protection is the setup the page is built to surface.

EPV Growth Premium — adding a franchise-value layer for compounders

Pure Greenwald EPV assumes "this company never grows again." That's fine for KO/JNJ-style mature steady names. For high-ROIC compounders like AAPL/MSFT/COST it's too pessimistic — AAPL pure EPV prints $84 against a market price of $283. Pure EPV becomes a floor-below-the-floor.

Greenwald's framework actually has three layers, not one:

1. Asset Reproduction Value — replacement cost, the absolute floor
2. Earnings Power Value — zero-growth perpetuity (our pure EPV)
3. Franchise Value — premium added on top when ROIC > WACC and the moat is verifiable

Until now we only computed the second layer. The new EPV Growth Premium scenario card adds the third layer:

EPV Growth Premium = EPV × (1 + (ROIC − WACC) × N)

where N is "moat persistence years" selected from ROIC magnitude:

• ROIC > 30% (exceptional compounder, AAPL services era) → N=7
• 20-30% (wide moat) → N=6
• 15-20% (above-average) → N=5
• 10-15% (competitive) → N=4
• < 10% (no clear moat) → N=3

Multiplier capped at 3.0× (prevents extreme ROIC from inflating beyond reasonable franchise range). When ROIC ≤ WACC, multiplier = 1.0× (the company isn't earning excess returns, so no premium applies).

How this number is used

Does NOT enter the ai_floor election. It's display-only as a scenario card, not part of the median election that produces the final recommendation. Reason: franchise value is a heuristic, not deterministic hard-logic — the linear 1 + excess × N form approximates Greenwald's true formula (1 − WACC/ROIC) × EPV × growth_factor closely in the 10-30% ROIC band but isn't the real article. We observe whether numbers feel right first, then decide whether to promote it into the recommendation pipeline.

How to read divergences

If a stock's EPV Growth Premium is well above pure EPV but well below market price, the engine is saying "market is paying for the franchise but hasn't bubbled" — could be a buyzone entry. If the two EPV values are close (ROIC ≈ WACC, no real moat), the franchise lens doesn't save you; rely on the other two floors. If EPV Growth Premium is missing ("—"), ROIC data is unavailable (5y EBIT incomplete or unusual capital structure).