Spectral Brand Theory — Framework & Toolkit

Brands are stellar objects. You are the observer.

graph LR
    subgraph Brand["Brand Signal Source"]
        direction TB
        SEM((Semiotic))
        NAR((Narrative))
        IDE((Ideological))
        EXP((Experiential))
        SOC((Social))
        ECO((Economic))
        CUL((Cultural))
        TMP((Temporal))
    end

    subgraph Pipeline["Perception Pipeline"]
        direction TB
        OSP[Observer<br/>Spectral Profile]
        PC[Perception<br/>Cloud]
        CC[Conviction<br/>Collapse]
    end

    Brand -->|"8 dimensions"| OSP
    OSP -->|"filter + weight"| PC
    PC -->|"threshold"| CC

    style SEM fill:#fce4ec,stroke:#c62828
    style NAR fill:#e8eaf6,stroke:#283593
    style IDE fill:#e0f2f1,stroke:#004d40
    style EXP fill:#fff8e1,stroke:#f57f17
    style SOC fill:#f3e5f5,stroke:#6a1b9a
    style ECO fill:#e8f5e9,stroke:#2e7d32
    style CUL fill:#fbe9e7,stroke:#bf360c
    style TMP fill:#e3f2fd,stroke:#0d47a1
    style OSP fill:#e8eaf6,stroke:#333
    style PC fill:#fff3cd,stroke:#333
    style CC fill:#d4edda,stroke:#333

What Is This?

Aaker drew the anatomy chart. SBT built the microscope. Same brands, higher resolution.

Spectral Brand Theory (SBT) models brands as multi-dimensional signal sources perceived differently by every observer. There is no "brand in itself" — only signals and observers. This toolkit provides a seven-module AI-native analytical pipeline that turns any capable LLM into a brand perception X-ray machine, producing structural diagnosis that traditional audits cannot.

SBT's contribution is not the eight dimensions themselves --- it is the measurement infrastructure: the observer-cohort framework, the spectral weight formalism, the metamerism detection mathematics, and the Brand Function specification. This infrastructure enables geometric positioning in perception space regardless of which dimensional decomposition is used.

Quick Start

Analyze your first brand in 10 minutes:

1. Copy the prompt from prompts/01_BRAND_DECOMPOSITION.md
2. Paste into Claude, GPT-4, or any capable LLM with your brand name
3. Get structured YAML output: signal inventory across 8 dimensions
4. Run all 7 modules for a complete Spectral Brand Audit

The 7 Modules

#	Module	What It Does	Input	Output
1	Brand Decomposition	Inventory signals across 8 dimensions	Brand name + materials	Signal map, D/A ratio
2	Observer Mapping	Define audience spectral profiles	Brand context	3-5 cohort profiles
3	Cloud Prediction	Predict per-cohort perception clouds	Modules 1+2	Cloud map per cohort
4	Coherence Audit	Score brand coherence (7 metrics)	Module 3	Grade (A+ to F) + type
5	Emission Strategy	Design dimensionally specific action plan	Modules 1+2+4	Strategy per cohort
6	Re-collapse Simulation	Test disruption resilience	Modules 3+4	Resilience profile
7	Resource Allocation	Optimal investment, alignment gaps, blind spots	Modules 1+2	Allocation plan + gap table

Each module has a prompt + YAML template in templates/.

Module 7 adds four practitioner features not present in Modules 1-6:

• Financial report input: Upload a P&L, 10-K, or departmental budget. The LLM maps line items to 8 dimensions using the Dimension Glossary.

• Code Execution Mode: When the LLM has a Python sandbox (Claude tools, GPT Code Interpreter, any agent), it runs validate_resource_allocation() directly — the math is computed by deterministic code, not LLM inference. Zero hallucination on the hard numbers.

• Data quality gates: Every output is tagged with its data source (survey, financial_report, llm_estimate). Estimated inputs trigger a mandatory warning: results are indicative only until validated with real data. If >30% of budget cannot be mapped to dimensions, the tool refuses to produce results rather than guess.

• Clarification protocol: When budget items are ambiguous (e.g., "marketing" spans Narrative, Semiotic, Social), the LLM asks the user to clarify the split rather than estimating silently.

• Balanced dimension ordering (dimension_order): PRISM-B elicitation experiments found a primacy effect in JSON-based prompts (eta-sq=.217 — dimensions listed first receive ~15.4% weight vs ~8.0% for last). The generate_prism_b_prompt() function supports three modes to counterbalance this bias:

  Mode	Behavior
  "canonical" (default)	Fixed S-N-I-E-So-Ec-C-T order. Backward-compatible.
  "latin_square"	Cycles through 8 orderings from a cyclic Latin square. Each dimension appears in every position exactly once across the 8-row cycle.
  "random"	Random permutation per call. Pass seed= for reproducibility.

This is the beginning of the README. Read the rest at the original. Read the full README ↗