// failure intelligence · how it works
The only system that matches a deal to documented failure patterns — with verified accuracy.
Matched against 5,032 documented autopsies, editorially curated. 85.9% Top-3 match accuracy (76.4% primary cause only) · the largest curated startup-failure dataset. It surfaces structural risks worth investigating — not a prediction of any one company's fate. CB Insights and PitchBook don't publish cause-match accuracy because they map outcomes, not structural precedent.
// data quality
What "editorially reviewed" means.
Every case in the database passes a structured review before it is published. The criteria are the same regardless of sector, geography or funding size.
The company must have demonstrably ceased operations, been liquidated, declared bankruptcy, or been acquired under distress. Pivots and rebrands are excluded.
The primary cause of failure must be identifiable from public documentation — founder post-mortems, regulatory filings, press coverage, court records, or investor statements. If the cause cannot be substantiated, it is left unassigned.
Key milestones — founding, peak, first warning signals, shutdown — must be reconstructible from at least two independent sources. Survival months are computed from these dates, not estimated.
Analytical fields — collapse style, hype cycle, moat type, fatal mistake, archetype — are assigned by a human reviewer, not generated automatically. Each field is a considered judgment, not a classification label from a model.
Cases that do not meet all four criteria are either excluded entirely or published with a "unverified" flag. The verified / unverified split is reflected in the precision metrics above — backtesting is always run on verified cases only.
// precision metrics
Model accuracy evaluation.
5-fold stratified cross-validation on the full editorial corpus. Every verified case is evaluated exactly once. n=5,032. Backtesting runs on verified editorial cases only.
| Metric | Accuracy | n | CI |
|---|---|---|---|
| Top-3 (primary or secondary cause) | 85.9% | 5,032 | ±1.2pp |
| Top-3 (primary cause only) | 76.4% | 5,032 | ±1.2pp |
| Version | Top-3 | Dataset | Note |
|---|---|---|---|
| V2 | 60.0% | ~1,500 | Baseline |
| V5 | 80.4% | 5,471 | Fatal mistake signal added |
| V6 | 61.9% | 16,112 | Dataset ×3 — retraining drop |
| V9 | 86.1%† | 17,044 | Multi-label metric only† |
| V12 (current) | 76.4% / 85.9%* | 5,032 | NB ensemble (α=0.8) + v8 retrieval |
* V12: 76.4% primary cause only / 85.9% primary or secondary cause. 5-fold stratified cross-validation, n=5,032. NB ensemble (α=0.8) blended with v8 structural retrieval (α=0.2). Intermediate versions (V3, V4, V7, V8, V10, V11) not listed. † V9: multi-label metric only (primary or secondary cause). V6: dataset tripled (5k→16k) with external data; retraining from broader corpus temporarily reduced precision.
// model pipeline
Five steps in the risk analysis.
Each analysis runs the same deterministic pipeline. Every score is traceable to concrete signals.
Exact parameters, bonus functions and bayesian adjustment values are proprietary and not published. The pipeline is auditable — every score is traceable to concrete signals.
// model signals
The six signals that most distinguish failures.
Approximate weights based on backtesting. A profile accepts up to 15 input fields; the model condenses them into 12 weighted scoring signals.
Primary pool filter. Highest discriminative power. Model anchor.
Exact or regional match. Captures regulatory and macroeconomic context.
The single field that most distinguishes failures in the data: +14.8% match accuracy when present.
+9.3% — founder profile correlates strongly with collapse type in the matched data.
Market cycle at founding. Neutralises temporal penalties.
New in V5. Exact and text-similarity matching of the primary fatal mistake. Includes unit-economics sub-classification (burn rate, CAC/LTV, margin compression).
Exact weights and internal model parameters are proprietary.
// reading the results
What the numbers actually mean.
The output is not a grade. Each number has a precise technical meaning that determines how to use it.
A score of 72/100 means the startup shares 72% of structural patterns with companies that failed in this pool, not that it has a 72% chance of failing. The model identifies resemblance, not destiny.
Structural similarityThe model identifies the primary failure cause in the Top-3 in 76.4% of analyses (85.9% when secondary cause also counts). Collapse causes overlap — if the cause you suspect appears in the Top-3, the signal is strong.
n=5,032Score ≥ 65: strong match, review the autopsy in detail. Score 40–64: moderate, relevant patterns but structural differences exist. Score 30–39: weak, pool is small, treat with caution. Closer score = more weight on that collapse timeline.
Min threshold: 30// common questions
Frequently asked questions.
The questions we get most from investors and analysts before they start using the platform.
// GET STARTED
Does any startup in your portfolio follow these patterns?
1 free risk match, no card — in 30 seconds. The full platform (IC memos, deal scouting, portfolio signals) includes a 7-day trial.