MODULE 2 · LESSON 5 / 5
Decision Engine™ — thresholds and verdicts
9 min read
From 4 scores to 1 verdict
The Decision Engine™ combines all four proprietary scores into a single Decision Score /100, then maps it to an actionable verdict. No verdict is ever produced without a confidence context.
Momentum Index™ × 0.35
+ Lifecycle Position™ (weight) × 0.25
+ Intelligence Grade™ × 0.25
+ Openness Score™ × 0.15
SOURCE: ground_truth.json → decision_engine.formula_weights
The 4 verdict thresholds
Verdicts are determined by multi-criteria AND logic — all conditions must be met simultaneously. A high score in one dimension alone is not sufficient.
INVEST
Momentum ≥70 AND Openness ≥40 AND Grade ≥70
Strategic window open — engage now
MONITOR
Momentum ≥45 AND Openness ≥25 AND Grade ≥50
Promising trajectory — premature to commit
EXPLORE
Momentum ≥25
Interesting signal — investigate before deciding
AVOID
Below all thresholds
Weak signal — limited opportunity
Blocking guards — algorithmic safety
Even if the Decision Score formula produces a high number, blocking guards can override the verdict when data quality or market structure indicates the signal is unreliable at that level.
BLOCKING GUARDS
intelligence_grade < 40→INVEST → EXPLOREpatent_count < 50→INVEST/MONITOR → EXPLOREstage == mature AND momentum < 40→→ AVOIDopenness < 20 AND verdict == INVEST→INVEST → MONITORstage == research AND verdict == INVEST→INVEST → MONITORWhy guards exist
A verdict without confidence context is dangerous. INVEST on a dataset of 30 patents with Intelligence Grade™ LOW is misleading. Guards ensure that verdicts match the actual reliability of the underlying signal.
Layer C — Tier gate (Phase 5.3)
Beyond Layer A (the AND-logic verdict) and Layer B (blocking guards), Layer C subordinates the user-facing verdict to evidence_certainty — a categorical reliability label (HIGH / MODERATE / LOW / VERY_LOW) derived from the Intelligence Grade™ framework. A verdict is never presented without a coherent confidence signal attached. The mapping is categorical, not threshold-based.
Why it exists: prior to its introduction, a verdict like « INVEST + Confidence 50 » was structurally possible because the AND-logic operates on raw scores while evidence_certainty derives from orthogonal signals (critical issues, silhouette, source coverage). Layer C subordinates the user-facing verdict to the certainty label so every output is auditable.
EVIDENCE_CERTAINTY → TIER → VERDICT OUTPUT
TIER_LOW — Directional signal labels
When evidence_certainty is LOW, the categorical verdict is mapped to a directional label that signals direction without committing to a categorical recommendation — the audit-grade language defendable in front of an IP firm’s clients. At MODERATE, by contrast, the categorical verdict is preserved as-is (with a simple confirm-first caveat); the raw verdict never reappears below MODERATE.
INVEST→OPPORTUNITY_SIGNALMONITOR→MIXED_SIGNALEXPLORE→WEAK_SIGNALAVOID→NEGATIVE_SIGNALTIER_LOW — Conditional directional signal
When evidence_certainty is LOW, the dataset is rich enough to compute every score, but Signal Integrity™ warnings limit the audit-grade confidence. The categorical verdict is mapped to a directional signal and surfaced as {signal} — Conditional in amber (for example MIXED_SIGNAL — Conditional). The signal is directional — resolve the flagged Signal Integrity warnings to lift certainty before any strategic commitment.
TIER_VERY_LOW — Verdict withheld
When evidence_certainty is VERY_LOW, the input itself is genuinely unusable (typically n < 10 patents, temporal coverage < 1 year, or structurally corrupted). The verdict is fully withheld and replaced by INSUFFICIENT_DATA — no signal can be defended. Distinct from Conditional: Conditional means « we have a verdict but warnings »; INSUFFICIENT_DATA means « the dataset cannot support any verdict ».
A confident verdict can still carry a caveat
The tier gate above moves the verdict when certainty drops. But some audit flags are grade-neutral: they leave the verdict and the confidence untouched, because the data is sound. Only its reading would be incoherent. CE6 is the clearest case: the Momentum Index™ can stay HIGH on a strong multi-year trend while the most recent filing year is already declining. The verdict can legitimately remain INVEST at HIGH certainty — but printing « move fast » next to that contradiction would be dishonest. So Stratensight keeps the verdict and the confidence, then surfaces the contradiction in the « Why this decision » and tempers the recommended action to « strong signal, but reconcile the flagged contradiction before committing ». A confident verdict and a « reconcile first » caveat can coexist, by design: the page never contradicts its own audit.
No verdict without confidence
Every verdict in Stratensight is accompanied by a confidence indicator. This is a core design principle: a verdict tells you what to do; the confidence tells you how much to trust it.
The confidence is derived from the Intelligence Grade™, dataset volume, and the number of blocking guards that were triggered. If multiple guards fire, the confidence drops even if the raw Decision Score is high.
✓
Module 2 complete
You now master all five Stratensightscores: Momentum Index™, Lifecycle Position™, Openness Score™, Intelligence Grade™, and the Decision Engine™.
Key takeaways
- Formula (indicative): Momentum×0.35 + Lifecycle×0.25 + Grade×0.25 + Openness×0.15
- INVEST requires 3 simultaneous conditions: Momentum ≥70, Openness ≥40, Grade ≥70
- Blocking guards override verdicts when data quality, volume, or concentration thresholds are violated
- Layer C subordinates the user-facing verdict to evidence_certainty — no output is ever produced without a coherent confidence band attached (TIER_HIGH preserves the categorical verdict, TIER_MODERATE preserves the categorical verdict + confirm-first caveat, TIER_LOW → directional signal '{signal} — Conditional' amber, TIER_VERY_LOW → INSUFFICIENT_DATA)
- No verdict is ever produced without a confidence score — the confidence tells you how much to trust the verdict
QUIZ DE VALIDATION
1. In the Decision Score formula, which component has the highest weight?
2. Momentum=80, Openness=55, Intelligence Grade=75 — which verdict?
3. What happens when Intelligence Grade™ < 40 and the calculated verdict is INVEST?
4. A dataset of 45 patents produces a verdict MONITOR. What happens?
5. Why does Stratensight never produce a verdict without a confidence score?
6. Openness Score™ of 15 combined with verdict INVEST triggers which guard?
7. Which lifecycle phase receives the highest weight in the Decision Score?
8. What is the purpose of blocking guards in the Decision Engine™?
9. Momentum=80, Openness=60, Intelligence Grade=75 — all three are above their INVEST thresholds. But Openness=30 instead — what happens?
10. mature lifecycle + Momentum Index™ LOW (< 45) produces which forced verdict?
11. Verdict = INVEST with evidence_certainty = MODERATE — what does the user-facing surface display?
12. evidence_certainty = VERY_LOW — what does the user-facing surface display, regardless of the raw verdict?