When NOT to Trust the Grade

Every scoring model has blind spots. Ours are documented here — because a tool you can't see the edges of is a tool you'll eventually cut yourself on

TRUST CALIBRATION

rendering diagram…

flowchart TB
    G[A GMI letter grade] --> STRONG[Strongest testimony]
    G --> WEAK[Weakest testimony]
    STRONG --> S1[Established, profitable,<br/>well-covered businesses]
    STRONG --> S2[Comparisons WITHIN<br/>a sector]
    STRONG --> S3[Trend across refreshes:<br/>grade decay is a real signal]
    WEAK --> W1[Turnarounds mid-inflection<br/>numbers lag reality]
    WEAK --> W2[Pre-profit growth stories<br/>check the cash-positive badge]
    WEAK --> W3[Banks/insurers/utilities<br/>leverage norms differ]
    WEAK --> W4[Thin-coverage small caps<br/>fewer factors testify]
    S3 --> USE[Use as TRIAGE:<br/>shortlist for human + AI<br/>investigation]
    W4 --> USE

The grade is a tip-off, not a verdict: strongest on established, well-covered businesses; weakest exactly where fundamentals lag the story. Calibrate accordingly.

The Known Blind Spots

Turnarounds look worse than they are. The grade scores trailing fundamentals. A genuine turnaround — new management, restructuring, an inflection just starting — shows up in the numbers after it shows up in reality. A D-grade company mid-turnaround is exactly the case the model is slowest to recognize.
Pre-profit growth companies get punished. Negative net income and negative ROE cost points regardless of why. A deliberate reinvestment story (the Amazon playbook) scores like distress. The cash-positive-despite-losses badge exists precisely to flag this pattern — read it.
Sector norms aren't normalized. The debt-to-equity bands treat leverage uniformly, but banks, insurers, and utilities run structurally high leverage by design. Financials routinely score lower than their sector reality warrants. Compare within a sector, not across.
Thin coverage means thin grades. Small caps and recent IPOs with sparse analyst coverage or short filing histories are graded on less evidence. The breakdown shows which factors were data-starved — a B built on three factors is weaker testimony than a B built on six.
The grade ages. Fundamentals refresh on filing cadence, not tick-by-tick. A grade computed before an earnings surprise doesn't know about it — which is why every grade carries its date, and why a months-old shared report warns you it's stale.
Analyst inputs import analyst bias. One factor (up to ±10 of 100) leans on price targets — a known-optimistic data source. It's deliberately the smallest voice in the room, but it's in the room.

What an A− Does Not Mean

A high grade means the company's trailing fundamentals are strong by the published rubric — nothing more. It is not a price call (a wonderful business can be a terrible buy at the wrong price), not a timing signal, not a prediction of next quarter, and not personalized advice. The grade deliberately excludes valuation-versus-price judgment beyond the analyst-target factor: fair-value math (like the P/E-compression quarters metric) lives alongside the grade so quality and price stay separate questions — the way they should be.

How to Use the Grade Well

Treat grades as a triage instrument: they compress hundreds of holdings into a shortlist that deserves human attention — the D's and F's that decayed, the surprising A's, the disagreements between the deterministic grade and the ML health score. Then investigate the shortlist with the hover breakdown (see which factors moved), the financial statements one click deeper, Ask Claude on the actual numbers, and your own research. That loop — grades tip off, humans and AI investigate — is exactly how the founder runs his own portfolio.

The standing disclosure, plainly: Grade My Investments is an analysis instrument, not a registered investment advisor. Grades inform decisions; they must not make them. Anyone selling you certainty about markets is selling you something else — we publish our blind spots instead.