The rubric, v0.1

Every weight, published.

A ranking you cannot inspect is a ranking you cannot argue with, and one you cannot take anywhere else. This is the whole rubric. It is computed entirely from free public sources: the EDGAR submissions API, the XBRL frames and companyfacts APIs, EDGAR full-text search, and daily indexes.

Provisional

This rubric has not been calibrated against any investor's revealed preferences. It was written from financing mechanics. It is deliberately instrument-neutral, which means it weights leading equity, writing converts, accumulating then negotiating, and taking down shelf paper roughly equally, and is therefore optimal for none of them. Calibration against thirty to fifty names you have actually financed is what fixes that, and it is the first thing an engagement does.

The universe

Public float$15M to $400M
SectorsLife Sciences, Technology, Natural Resources
Sourcedei:EntityPublicFloat, swept across nine quarterly instant frames and unioned by CIK
Size as built676 issuers · Life Sciences 453, Technology 158, Natural Resources 65

Why float rather than market cap. EDGAR carries no market capitalization and every source that does is paid. Float is free, arrives in bulk, and is the exact input the Form S-3 baby shelf limit keys off, so shelf-capacity arithmetic comes free. Float runs below market cap, so the $15M to $400M band approximates a $20M to $500M market cap target.

Why the frame sweep. Float is reported as of the last business day of the second fiscal quarter, so it scatters across instant frames by fiscal year end. Calendar-year filers land in Q2 frames of roughly 4,300 issuers each; off-cycle filers land elsewhere in frames of 250 to 500. Querying a single frame returns 260 issuers and a broken universe.

The hard gate

A name is scored only if at least one of these fired. No trigger, no score.

Deliberately not a trigger: the mere presence of an annual proxy. Every company files one, so gating on it passed 41 of 42 issuers on the first live run and made the gate meaningless.

Score, 0 to 100

1. Runway urgency, max 30

Cash and equivalents against trailing operating cash burn, both from XBRL, expressed as months remaining.

RunwayPoints
Under 3 months30
3 to 625
6 to 918
9 to 1212
12 to 185
Over 180

Positive operating cash flow means no burn-based runway applies and this component scores zero. Work out a band →

2. Structural need for equity, max 25

SignalPointsWhy
Listing deficiency against a stockholders' equity standard25The company cannot cure this with debt, and it runs against a compliance-plan deadline. A structural buyer of exactly this kind of paper, on a public clock.
Listing deficiency, other cause (bid price, market value)12Weaker. A reverse split or time can fix it.
Going concern, substantial doubt15Formulaic enough to detect reliably, serious enough that a board has already had the conversation.
Reverse split proposed in a DEF 14A10Usually runs ahead of or alongside a raise. Requires text confirmation, not just the presence of a proxy.

3. Readiness to transact, max 20

SignalPoints
Fresh S-3 inside 90 days15
Baby shelf constrained, float under $75M12
Active ATM, 424B inside 180 days10
Prior unregistered sale, 8-K Item 3.02, inside 24 months8

Baby shelf scores as a positive. Under Form S-3 General Instruction I.B.6, an issuer with public float below $75M cannot sell more than one third of that float in primary offerings over any twelve months. The shelf structurally cannot carry a full raise, so the balance has to come from a private placement. That is the opening. 323 of the 676 issuers in the universe sit under this cap. Calculate a specific issuer →

Prior Item 3.02 filings do double duty: they prove the company will do this kind of deal, and the exhibit names the placement agent who ran the last one.

4. Capital structure pressure, max 15

SignalPoints
8-K Item 2.04 acceleration event15
Convertible maturing inside 12 months15
Variable-price convertible outstanding12

A variable-price convert is a company with an obvious interest in being refinanced out, and a clear reason to talk to someone who can do it.

5. Distress and governance, max 10, capped

SignalPoints
NT 10-K or NT 10-Q late filing5
Auditor change or resignation, 8-K Item 4.015
Officer or director departure, 8-K Item 5.023

None of these is a financing signal alone. Stacked on a runway number, they change the read.

6. Sector timing bonus, max 10

SectorSignalPoints
Life SciencesA readout or PDUFA date inside 9 months that current runway does not reach10
Natural ResourcesAn S-K 1300 or NI 43-101 milestone whose next stage needs capex8

Life Sciences scores higher because the date is public and the arithmetic is unambiguous.

Exclusions and penalties

RuleEffect
Float outside $15M to $400MExcluded from the universe
Float under $10MExcluded, too small to place
Over 24 months runway with no listing or debt triggerExcluded
Operating cash flow positive with no financing triggerExcluded
Financing announced within 60 daysminus 20, they just did a deal

Staleness guards

The failure mode this product actually has is not a wrong weight. It is a confident number computed from a stale input. One issuer once ranked top ten on a runway of 0.9 months derived from cash dated 1,336 days earlier. The arithmetic was correct and the answer was worthless.

GuardBehaviour
Cash older than 270 daysRunway urgency scores zero and the table prints n/a. The zero is pushed explicitly with its reason, because a silently missing component looks like a bug while a printed zero is an audit trail.
Float older than 400 daysThe as-of date is printed next to the number. 20% of the universe is in this state.

Thresholds

What the model does, and does not, do

The deterministic layer computes roughly 90% of this rubric with no language model at all: runway arithmetic from XBRL, 8-K item codes from the structured submissions index, form-type matching, and float arithmetic.

A single model call per candidate handles only what genuinely lives in prose: going-concern language, which specific listing standard was breached, convertible terms, and dated sector catalysts. It also writes the dossier. It is given no tools and no ability to search, because every fact it needs is supplied in the prompt and anything it could reach for is a fabrication risk.

The system ranks and explains. A person decides. Nothing here recommends, sizes, or times anything, and nothing here is investment advice.

Known limitations of v0.1

  1. Not calibrated. Weights are reasoned from financing mechanics, not fitted to anyone's revealed preferences.
  2. Float is a lagging proxy for market cap. Measured at fiscal Q2, and materially stale for an issuer whose price has moved.
  3. Instrument-neutral. The ranking should reorder depending on whether you lead equity PIPEs, write convertible debt, accumulate then negotiate, or take down shelf paper. Until that is answered it weights all four roughly equally.
  4. Precision is unmeasured. The backtest establishes recall only.
See what it would have caught