Feature — Confidence Score

Know when to trust the diagnostic.

Every episode carries a 0–100 Confidence Score — a transparent rating of how well-supported the diagnosis is. High confidence means act now. Developing means validate one variable first. The score is built into every episode before you commit to implementing anything.

78 High Confidence
Diagnostic supported by strong signal pattern — act on this episode's framework.
68–88
Score range across
all 46 current episodes
11
High-confidence episodes
scoring 80+
42–65%
Customer journeys invisible
to standard attribution
Berman & Katona, 2024 · Marketing Science
$72B
Ad spend wasted on
invalid traffic (2024)
Statista / Lunio, 2024
Score tiers — what each range means
80+
High
Strong signal — act on this
The diagnostic pattern has appeared consistently across multiple independent industry contexts. Root cause identification is reliable. All 11 episodes scoring 80+ are corroborated across at least 4 industry contexts with peer-reviewed source backing. Implement the episode's framework with confidence.
Recommended action
Apply the episode's framework directly to your current situation.
60–79
Developing
Good signal — validate one variable
The pattern is present but has meaningful variation across contexts. The core diagnosis is likely correct but one factor may shift the recommendation. 35 of 46 episodes sit in this band — it reflects honest calibration, not low quality.
Recommended action
Review the episode's Confidence Note — it flags the specific variable to check before acting.
<60
Emerging
Weak signal — form hypothesis, collect data first
The pattern exists but the causal pathway is contested or context-dependent. Useful for framing the problem — not for immediate implementation. No current episodes score below 60; this band is documented for completeness.
Recommended action
Use the episode to form a diagnostic hypothesis, then collect the data points it identifies before implementing.
Why calibrated uncertainty matters — the research
Rodman et al., 2023 · JAMA Network Open
Clinicians and AI diverge most when context requires reframing — not just pattern matching.
Rodman et al. found that the performance gap between expert and AI is largest when new information requires updating the problem frame itself. Experts revised their assessments based on context and case framing. The Confidence Score operationalizes this: it tells you how stable the diagnosis is across varied contexts before you act.
Rodman et al. AI vs Clinician Performance. JAMA Network Open, 6(12):e2347075, 2023.
Berman & Katona, 2024 · Marketing Science
42–65%
Customer journeys now invisible to standard attribution — making diagnostic confidence structurally essential.
Following iOS privacy changes and cookie deprecation, a significant portion of the customer journey cannot be reliably attributed. When measurement systems are structurally incomplete, knowing the confidence level of a recommendation isn't optional — it's the only honest way to act.
Berman, R. & Katona, Z. Privacy changes and attribution model accuracy. Marketing Science, 2024.
Zhang et al., 2024 · Wharton School
23–31%
Average overestimation of marketing performance — the cost of acting on unqualified diagnostic outputs.
When metrics are accepted without cross-validation, performance is systematically overstated by nearly a quarter. The Confidence Score is the cross-validation signal built into every episode — it flags where the diagnostic has been stress-tested across contexts and where it hasn't.
Zhang et al. Channel silos and marketing performance overestimation. Journal of Marketing Analytics. Wharton, 2024.
Kassirer, 2025 · Annals of Internal Medicine
The expert advantage is problem construction — not just pattern selection.
Kassirer identified the structural limit of AI decision support: it can process supplied information quickly but cannot judge context, weight incomplete history, or resolve ambiguity. Under ambiguity, the human diagnostician actively constructs, revises, and tests the problem frame. The Confidence Score tells you when ambiguity is present and what to resolve before acting.
Kassirer, J.P. Artificial Intelligence in Medical Practice: Is It Ready? Annals of Internal Medicine, 178(4):596–597, 2025.

"AI can process a dashboard. It cannot recognize when the dashboard is measuring the wrong thing. That judgment — knowing when to trust the signal — is what no AI adoption makes redundant."

— Light2Path Research Paper R06, drawing on Kassirer, Annals of Internal Medicine, 2025
How the score is calculated
Mechanism strength — how clearly the causal chain from symptom to distortion is established 35%
Dashboard invisibility — how completely the distortion is hidden in standard reporting 25%
Economic consequence — how computable and material the downstream impact is 20%
Cross-industry pattern — how many independent industry contexts confirm the finding 15%
Source integrity — independence of primary sources from vendor or commercial interest 5%