Sits above the per-competitor signal layer and produces market-level strategic intelligence — when to launch, how healthy your competitors are, and what your revenue risk is. Launch windows, moat health scores, causal outcome validation, and a 33-signal significance registry.
Composite 0–100 score across all tracked competitors. Category vacuum bonus of +20pts per additional weak competitor (up to ×3) — fires when ≥2 competitors show simultaneous weakness.
Outputs: driving_competitors list (top 5 by weight) · top_opportunity narrative · action_steps list (up to 5, urgency ranked 1–10) · window_days_left estimate.
Two-layer revenue model: a Shopify-anchored calibration engine and a per-competitor estimator. Low / mid / high range output — never a single false number.
Shopify-anchored: syncs against your own Shopify revenue actuals. Detects calibration drift and corrects revenue estimates when the model diverges from real data. Revenue tier auto-updates based on observed actuals — your playbook tier stays accurate without manual input.
Estimates competitor revenue from product count, price points, restock frequency, and catalog velocity. Per-competitor revenue estimates stored and tracked over time. Three-band output: low / mid / high — never a single false number.
Every prediction is tracked to resolution. The system knows if it was right, wrong, or inconclusive — and self-corrects accordingly.
Compares prediction accuracy against what would have happened by chance. A prediction that's right 70% of the time when the base rate is 65% is nearly worthless — the system tracks this delta, not raw accuracy.
Flags predictions where counter-evidence accumulates. If a launch prediction fires but no inventory shipments, trademark filings, or ad bursts follow within the window, the prediction is actively marked against.
Predictions that can't be confirmed or denied expire as expired_unconfirmed — not noise, not false positives. Preserves data quality for accuracy trend analysis.
Dashboard component showing validated accuracy percentage broken down by prediction type. signal_patterns table populated for accuracy trend analysis over time.
0–10 weighted average across 4 components. Trend direction tracked as improving / stable / declining (−1 to +1 scale). Feeds directly into Prediction Engine confidence tier adjustments.
8 niche context overlays that override or augment the base Playbook Engine entry. Each niche carries category-specific benchmarks, CAC data, and margin targets.
Monitors US import records from Panjiva. Classifies each shipment into one of 4 categories and corroborates product launch and inventory stress predictions.
Large shipment correlated with imminent product launch.
Standard inventory replenishment — no launch signal.
New HS codes indicating new category entry.
Routine import with no corroborating signals.
significance-config.ts — 33 signal types, each with a calibrated significance score (0.00–0.92). Baselines are always 0.00: a first sighting is not an event. Scores drive the urgency router and compound rule weighting.
Beta is launching soon. Save your spot on the waitlist — or apply as a beta tester and get early access plus exclusive perks.