WickedForGood Methodology.pdf

Wicked: For Good — Methodology Overview

A three-layer forecasting system combining historical comps, AI visibility signals, and ensemble modeling.

AI visibility signals consistently move 7–12 days before traditional tracking, enabling earlier detection of demand shifts.

Three-Layer Forecasting Framework

Historical Performance Models

• Normalized performance curves
• Domestic ↔ global ratios
• Holiday corridor uplift
• Trailer shock elasticity
• Release pattern seasonality

AI Visibility Metrics

• SoP (Share-of-Prompt)
• AI Signals
• Prompt
• Density
• Sentiment
• Media reference velocity
• Cross-model consistency checks
• No search, no presales curves, no exhibitor data, no leaks, no surveys, no

Ensemble Forecasting Engine

• Multi-model ensemble tuned for AI visibility
• Cross-signal weighting & stability controls
• Temporal calibration for release cycles
• Confidence interval calibration

Forecasting Pipeline

Historical Comps

300+ theatrical titles

AI Visibility Signals

LLM visibility

Ensemble Model

Multi-model blend

Forecast Output

emberos projected $XXM

Each stage feeds normalized features into the multi-model ensemble for unified forecast output.

Data Input Layers

Visibility Layer

SoP, SSI, PDI

Context Layer

Stable market factors that influence baseline demand before AI visibility signals are applied.

Release Landscape

! These inputs act as normalization anchors so the model doesn't overreact to short-term AI or buzz fluctuations.

Stability / Variance

External factors such as reviews, competition, and weather introduce variance not captured directly within AI visibility signals. Reach out to sales@emberos.ai if you'd like to deeper into this.

• Forecast lock #1: Nov 14, 2025, 12PM PT
• Forecast lock #2: Nov 20, 2025, 12PM PT
• No model edits post-freeze
• Outputs timestamped + archived

This protocol prevents any modification to the forecast after freeze, ensuring integrity and independent verification.