Diagnostic Dashboard
3-Agent Adversarial PipelineMRR, churn, burn, NPS, LTV:CAC, runway — the signals that predict failure 3–6 months early
Voyage AI embedding → Atlas Vector Search + BM25 RRF → Gemini Flash parallel scoring → top pattern selected
Second independent Gemini 3 Flash instance — actively seeks counter-evidence. >10pp divergence triggers DISPUTE and catches false positives.
Oracle Score (transparent formula) · Escape Plan (deterministic algebra) · Cascade Graph ($graphLookup) · Survival Playbook
Your Current Metrics
Enter your startup's numbers below. Or let Oracle extract them automatically from any text.
Proven on Real Failures
These are actual Oracle outputs — we ran the system against each company's representative pre-collapse metric profile. Click any card to reproduce the result yourself.
Pattern similarity scores shown above are live Oracle outputs — run by us before submission against each company's representative pre-collapse profile (illustrative figures chosen to capture the real failure pattern, not audited financials). Click any card to reproduce them on the live system. The Quibi result is projected from its early trajectory. Pattern similarity reflects narrative match against the failure library, not a statistical probability of failure.
A second independent Gemini 3 Flash instance re-evaluates the match with deliberate skepticism — actively looking for counter-evidence. If their confidence assessments diverge by >10 percentage points, a DISPUTE is triggered. This catches false positives a single agent would miss.
Strongest counter-evidence found by Challenger:
Startup Trajectory Forecast
Comparing your projected runway curve against typical matched failure profiles, showing the course-correction recovery path.
Your Metrics vs. Pattern Trigger Thresholds
| Metric | Your Value | Pattern Trigger |
|---|
Warning Signals Detected
Historical Outcomes
Estimated days to crisis:
Survival Playbook
Used by every startup that survived this pattern
Companies That Matched This Pattern
Cocktail Alert — 2 Co-occurring Failure Patterns
Two or more failure patterns active simultaneously. Risk compounds non-linearly — each additional pattern roughly halves the survival probability. Triggered when ≥2 patterns exceed 60% confidence.
max(2%, min(rates) × 0.5n−1) — compound survival: take the lowest individual survival rate and halve it for each additional co-occurring pattern (floor 2%). Example: 20% survival + 2nd pattern = 10%, + 3rd = 5%.
Your Escape Plan
Minimum metric changes to drop your pattern match below 60% (the alert threshold). Each intervention is exact: "reduce churn from X% to Y%." Ranked by confidence impact. Pure algebra on the pattern's trigger conditions — no AI, reproducible every time.
Estimates based on pattern trigger thresholds. Difficulty ratings: easy (<20% metric change needed), medium (20–45%), hard (>45%).
Failure Cascade Graph
Failure isn't a single event — it's a chain. Each node is a failure mode that triggers the next. Transition probabilities start as research estimates and self-improve via Bayesian blending as real-world cases are observed (p = 0.3×initial + 0.7×empirical). Graph traversal via MongoDB $graphLookup · written atomically in an ACID transaction · probabilities updated live via Change Streams.
No Dangerous Patterns Detected
Your current metrics don't match any of the 100 documented high-risk failure patterns. Your trajectory looks healthy for this stage.
Run this monthly. Failure patterns build over 3–6 months. Early detection is the difference between a pivot and a shutdown.
Metric Trend
Your tracked metrics over time — MRR growth rate vs. churn rate. Pattern detection dates shown as vertical markers.
Decision Auditor
Before you make a major decision, ask the Oracle. It will evaluate your decision against 100 documented failure patterns and tell you if founders in similar situations succeeded or failed — and why.
Try: "Should I hire 3 engineers this month?" · "Should I raise prices by 20%?" · "Should I expand to Europe?"
Oracle Pre-Mortem
Simulate how a strategic decision ripples through your metrics over 6 months. Gemini projects metric changes at months +1, +3, and +6 — Oracle Score at each horizon, plus which failure pattern your month-6 trajectory would trigger.
MongoDB pipeline: Gemini Flash projection → Oracle Score (deterministic) → Pattern match on month-6 metrics
Uses your last analysis metrics as the baseline. Run an analysis first if you haven't.
VC Portfolio Mode
Parallel Analysis · Up to 20 StartupsPaste your portfolio companies — the Oracle runs concurrent async analysis on all startups simultaneously and returns a risk-ranked heat-map. Uses MongoDB $facet for live cohort benchmarking per company.
Cohort Intelligence
MongoDB $bucket + $facet — see how your startup ranks against similar companies at the same stage. Where do you sit in the distribution of 302 startups the Oracle has analyzed?
Where your Oracle Score sits in the distribution of all analyzed startups in your industry and stage
Scores grouped into 10-point bands — see the full distribution shape, not just your rank
What the top-quartile startups in your cohort did differently — their avg churn, burn multiple, and NPS
—
Pattern Library
100 documented failure patterns from YC post-mortems, CB Insights, Sequoia, a16z, and public founder post-mortems.
Submit Pattern
Contribute a new failure pattern to the collective oracle intelligence and help other founders survive
Submit a Failure Pattern
Witnessed a startup failure pattern not in our library? Share it — reviewed submissions are added to the pattern library and help all founders. Your story could save someone else's company.
Include specific metrics if possible (churn rate, burn multiple, runway at the time). The more quantified, the more useful the pattern becomes for other founders.