BillionToOne [BLLN] — Criteria & Two-Horizon Valuation

Phase Space Research · 2026-07-29 · Framework: Criteria (references/criteria.md, 2026-07-29), superseding the gate framework v1.4.2 · Archetype: INFLECTION

This document supersedes the six-gate block in 03_Trade_Construction.md and the Gate 4 expected-return table in 02_Valuation_Analysis.md. The company research, payor-conversion analysis, ASP work, mention-frequency series and correlation matrix in those files stand unchanged.

There is no position verdict in this document. The previous manifest carried investment_decision: "Watchlist" with binding_gate: "2_variant_vs_consensus"; both fields are retired.

0. What moved, and why — this is the largest reversal in the batch

BLLN was the closest miss in the entire prior record — rejected on a 4.06pp shortfall against a cash hurdle that no longer exists, with a binding gate that no longer exists either. Under the current framework both BINDING Criteria pass.

1. Data hygiene — the automated scan was wrong by 4.1x, and it mattered enormously

1.1 Share count

Shares used: 45,995,484 — 41,442,834 Class A + 4,552,650 Class B, cover page of the Q1 2026 Form 10-Q, as of 2026-05-04.

The 2026-07-28 universe scan returned 11,178,467. That figure is us-gaap:CommonStockSharesOutstanding with a period end of 2025-09-30 — the pre-IPO Class A count, before the 2025-11-06 IPO converted the redeemable convertible preferred. BillionToOne's current cover-page count is tagged dimensionally per share class, so dei:EntityCommonStockSharesOutstanding returns HTTP 404 and the fallback reached a stale, pre-IPO tag.

The EntityPublicFloat sanity floor did not fire, because BLLN publishes no EntityPublicFloat tag at all — it IPO'd in November 2025 and has not yet filed a 10-K with a float measurement date. A check that depends on an optional tag is itself subject to D1.

The consequence was not cosmetic:

The old scan record would have produced a confident PASS for the wrong reason. coverage_scan.py's own comment warns of exactly this: "A wrong share count is worse than a missing one: it yields a confident valuation rather than an INDETERMINATE." The conclusion happens to survive the correction, but it survives on different arithmetic, and that is stated rather than glossed.

Diluted weighted-average shares for Q1'26 were 53,028,360; on that basis EV is $6,628m and the required CAGR rises to 55.0%, still a +48.4pp margin. The conclusion is not sensitive to the choice.

1.2 TTM revenue

The universe scan returned $289.293m — 18% low — because coverage_scan.ttm_revenue() sums the last four tagged quarterly periods without checking they are consecutive, and here summed 2024Q3 + 2025Q1 + 2025Q3 + 2026Q1. The same defect overstated SMR by 62% and understated BE by 36%.

Q1'26 revenue grew +83.8% year on year.

1.3 Net cash

$447.4m. At 2026-03-31: cash and equivalents $537.450m − long-term debt $90.005m. The scan used a stale FY2025 debt figure of $57.226m. The common stock warrant liability of $8.924m is not debt and is excluded.

1.4 The prior-period true-up — the existing memo's finding is confirmed in the filing

The memo flagged an accelerating prior-period revenue line. The Q1 2026 10-Q confirms it verbatim: "the Company recorded $9.2 million and $2.9 million, respectively, as revenue related to performance obligations satisfied in prior periods" for Q1'26 and Q1'25 — and states that revenue for services performed in FY2025 covered by newly contracted payors "contributed slightly more than half" of that $9.2m.

$9.2m is 8.5% of Q1'26 revenue. Excluding it, underlying Q1'26 revenue is $99.2m and year-on-year growth against Q1'25 as reported is +68.2% rather than +83.8%. Both figures are reported; neither is suppressed. This is the single most important qualifier on the Valuation Criteria PASS below.

2. The Criteria

2.1 Quality Criteria — INFLECTION standard

BLLN sits just above the archetype boundary — TTM operating margin 10.2% against the 10% COMPOUNDER threshold — and is classified INFLECTION on character: margin expanding violently, growth very high, value in the terminal period. As with BE, INFLECTION is the stricter test because it imposes a gross-margin level.

F-score and gross-profitability LEVELS are demoted to context per criteria.md. They happen to be strong anyway — accruals −1.14%, GP/TA 0.356 (top-decile), F-score 7/8, the only miss being equity issuance, which is mechanical in an IPO year. Unlike TXG and TWST, there is no quality-of-earnings finding that independently argues against the name.

The two genuine earnings-quality concerns are invisible to every one of those factors and are recorded here rather than in a scorecard that cannot see them: unremediated ICFR material weaknesses (scoped to financial instruments and ITGC, not revenue) and the prior-period true-up line of §1.4. This is calibration item S5's second instance — the scorecard measuring the wrong object.

2.2 Liquidity Criteria

PASS on equity, FAIL on options — and the options finding is exactly what this Criteria exists for.

• Equity: easy_to_borrow: true, 30-day ADV 703,753 shares / $83.0m, short interest 10.51%, 3.6 days to cover and declining.
• Options: the chain was pulled before any structure was proposed, per the HCA precedent. It exists — 232 contracts across four expiries, strikes $40–190 — but total chain volume is 1,585 contracts (~$21m notional) and median ATM bid/ask spreads run 32.3% of mid (Aug), 26.4% (Sep), 16.4% (Dec), 10.8% (Mar-27). A vehicle that cannot be filled is not a vehicle.
• Recorded because it is counter-intuitive and useful: ATM implied volatility of 72.6–76.8% sits BELOW realised volatility of 85.7%. Premium is not overpriced on this name. The spreads, not the pricing, are what make options unusable. IV was recomputed by Black-Scholes from mid prices rather than trusted from the API, and Alpaca's field checked out to within ~0.01 vol points.

3. Valuation Criteria — the two mandatory outputs

3.1 The implied-path test

What the price requires: a 50.3% revenue CAGR for five years.

Demonstrated: 103.4%. Margin = 103.4 − 50.3 = +53.1pp.

Demonstrated CAGR window: FY2023 $71.729m → TTM 2026-03-31 $354.537m, 2.25 years. Short, and stated as such — BLLN has only three fiscal years of filed revenue.

Implied compression, as a number. BLLN trades at 157.3x TTM EV/EBIT on a margin that has only just turned positive. The exit multiple is 33.0x — a compression of 124.3 turns, −79.0%. The price already embeds a very large multiple compression, which is the mechanical reason the required growth rate is so much lower than the demonstrated one.

A caveat on the anchor's identification. The growth-matched set has n = 7, only just above the minimum of 5. It is a thinner anchor than BE's (n=33) or SMR's (n=83) and the multiple should be treated accordingly — which is what the sensitivity below is for.

3.2 Sensitivity — over the exit multiple, never over scenario probabilities

This directly replaces the instrument the old memo used. That memo ran its range across the bear weight and reported a flip at a 22% bear weight versus the 30% used — while point-estimating the multiple. criteria.md names this failure explicitly on NTRA: the analysis was run on the parameter that could not change the answer.

The flip point is a 12.0x exit EV/EBIT. The exit multiple would have to fall to roughly one third of the growth-matched anchor — and to a level below every comparator in the set — before today's price stopped being justified by what BillionToOne has already demonstrated.

Valuation Criteria: PASS. The implied path sits materially below what the business has demonstrated, and it does so robustly across the entire multiple range tested.

The honest qualifier, stated with the PASS rather than after it. The demonstrated 103.4% CAGR is measured over 2.25 years from a small base, and 8.5% of the most recent quarter's revenue was a prior-period true-up. On the ex-true-up growth rate the margin narrows but does not close. The PASS is genuine; it is not a claim that 103% growth continues.

3.3 What the price actually requires, restated in business terms

At a 50.3% CAGR, revenue reaches roughly $2.72bn by 2031. The existing memo's own reverse-engineering reached a comparable place from the other direction — "FY2026 $462m growing at an initial 37.4% and fading 10% per year — reaching ~$3,515m" — and judged it demanding. The difference is that this test measures the requirement against BillionToOne's own demonstrated rate rather than against a cash hurdle, and on that comparison it is not demanding.

The substantive risk is unchanged and is not a multiple argument: the price requires an oncology franchise with 1 publication and 0 registered MRD trials to become a large share of the company. That is logged under Downside Criteria, where it belongs, rather than as a valuation veto.

3.4 The 12-month target

NTM revenue: $533.4m = 5/12 × FY2026E consensus $454.14m (8 analysts) + 7/12 × FY2027E consensus ~$590m. Both figures are from the existing valuation document. Company FY2026 guidance is $450–465m.

EV / NTM revenue today: 10.67x.

Own-multiple anchor and percentile. BLLN listed on 2025-11-06, so its own history is short. Over the 96 sessions since its first 10-K gave the market an as-known annual revenue figure, EV/Sales ranges:

BLLN trades at the 100th percentile of its own trading history — an all-time-high multiple.

Unlike BE and SMR, this anchor is usable: BLLN's growth regime did not change across the window, so the series is measuring the multiple rather than the revenue base. The window is short (96 sessions) and that is the caveat, not an identification failure.

12-month target: $152, +13.8% to spot. Band $132 – $181.

The mechanism is worth naming, because a target above spot from a name at its all-time-high multiple looks contradictory: it is not a re-rating call. It is NTM revenue growth (+50% on TTM) outrunning a multiple compression from 18.6x to its own 12.3x median. Growth wins; the target is above spot despite the multiple falling.

Sanity band. The previous house target was $136.01 (+1.8%), built from a probability-weighted scenario set. The new target is +13.8%. The difference is instrument, not view: a scenario weighting with a −44% bear at 30% drags the mean down mechanically, whereas the 12-month instrument asks what the name is likely to trade at. valuation.md records that the reference professional's longs sit at a +20% median to spot while 16 of 16 house targets sat below spot; +13.8% is the first output in this batch that sits inside the professional's observed range.

4. Downside Criteria — MEASURED, blocks nothing

Under the previous framework the bear case fed the E[R] arithmetic that produced the rejection. It no longer does. It is logged with its cause and probability, Brier-scored by ledger_scorer.py, and it constrains nothing. Inverse-volatility sizing is the active control — realised volatility 85.7%.

5. Momentum Criteria — MEASURED, entry timing only, and partly INDETERMINATE

Momentum governs when, never whether. All veto and blocking language is deleted.

This is a live instance of the D1 discipline on a name where the temptation to substitute is strong. A scanner that treated the absent 12-1 window as a zero would have placed BLLN near the universe median (+2.1%) and assigned it a middling quintile — a fabricated reading presented as a measurement. The correct output is INDETERMINATE, with the 6-1 measure reported alongside it because that one is computable.

On the computable evidence the tape is strongly with the name: 6-1 momentum at the 86.5th percentile and a close at 100.0% of the 52-week high. RSI-14 at 63.2 says today is not an obviously good entry — which is a statement about timing, and the only thing this Criteria is permitted to say.

6. What this document deliberately does not say

• No Long, Short, Watchlist or Avoid. The prior investment_decision: "Watchlist" and binding_gate: "2_variant_vs_consensus" are retired.
• No variant-versus-consensus test. Gate 2 is dissolved. The +1.8% FY2026 gap is reported under Consensus Criteria and blocks nothing — criteria.md: "A quota gap leaves this blank and blocks nothing."
• No E[R] versus a 4.7% cash hurdle, and therefore no "fails by 4.06pp".
• No claim that the oncology franchise is proven. It is not; that is the bear case.
• No options structure. The chain was pulled and it cannot be filled.

7. Provenance