When every signal that makes a stock look cheap is an artifact, the trap is in the screen

Claim

Before calling anything cheap, ask of each cheapness signal: would this reading exist if the price had not fallen? A signal that only exists because of the price fall carries no information about value — it is the price fall, restated.

The three artifact forms

Signal	Artifact mechanism	The check
Low trailing P/E	A non-cash or one-off credit sits in TTM earnings	Compare reported to adjusted earnings; a forward P/E ABOVE the trailing P/E on a company guiding down is the free tell
High dividend yield	Yield rose because price fell, not because the dividend grew	Check the dividend's own growth rate. Flat or cut + higher yield = price-caused
Large drawdown	The de-rate was correct — a premium was withdrawn	Ask what the premium was for, and whether the company still earns it

NVO had all three at once, and the third is the one that decided it: management itself guided 2026–30 growth to "in line with industry peers" and named the 2032 LOE. When a company formally surrenders its growth premium, the drawdown that removed that premium is not a discount — it is a repricing to the correct cohort. Buying the drawdown is buying the premium back at the moment it was disclaimed.

Why the conjunction matters more than any single flag

Each of these has its own note. The pattern worth naming is that they co-occur, because they share a cause: a de-rating produces all three mechanically. So the check is not "is one of these an artifact" but:

Count how many of the cheapness signals survive the artifact test. If the answer is zero, the stock is not cheap — the screen is broken for this name.

That is a stronger and faster test than valuing the business, and it is available before any modelling.

The corollary: this is not the same as a value trap

analysis_notes.md §0 defines a value trap as poor business + cheap price. An all-artifact cheapness profile is a different thing, and conflating them produces the wrong action:

Value trap — the cash is not real. Sell, or never buy.
All-artifact cheapness — the cash may be entirely real; the price is simply not low. Hold if owned, do not buy, and re-derive a level at which it would be cheap.

NVO is the second kind: FCF DKK 28.3B → guided 45–55B, first capex decline, buyback restarted 10x, payout 183% → ~104%. The cash is improving. It is fairly valued at ~11.9x adjusted trailing for a mid-single-digit grower. "A buy at $39.49 would be a value-trap buy; a buy at $31 on the rebuilt numbers would not be" — same company, same facts, different price.

The operational danger: stale entry zones

An entry zone derived from a pre-de-rate fair value will sit above the new fair value, so the artifact signals and the zone fire together and the system instructs a purchase. NVO's $33–40 zone was revoked on 2026-09-22 with spot at $39.49, roughly 10% above the rebuilt centre. See [[pitfall-stale-fair-value-is-most-costly-on-winners]] and [[pitfall-stale-entry-zone-suppresses-a-name]] — this is the third face of the same defect.

The rule

Run the three-way artifact test before any valuation work. It is nearly free.
A forward P/E above the trailing P/E on a company guiding down is a free red flag — it means the market already knows the trailing figure is inflated.
Decompose a yield into dividend growth and price change before calling it income.
When a management team publicly re-cohorts itself, re-derive the multiple from the new peer set — do not adjust the old multiple downward. Adjusting anchors; re-deriving does not.

What would falsify this

A name where all three signals are artifacts and the stock is nonetheless cheap on rebuilt numbers — entirely possible, since the test rules out evidence of cheapness, not cheapness itself. That is precisely why the rule ends in "re-derive a level," not "avoid."

History

2026-08-05 — NVO analysed; one artifact caught (the 340B provision reversal inflating the P/E) and recorded as "the cheapness is partly a data artifact."
2026-09-22 — the other two identified after the Capital Markets Day; entry zone revoked; pattern generalised from the conjunction.