Our Methodology

Buyology Labs does not simulate brain scans. We do not have access to your audience's brain activity. What we do is predict neural response patterns using a scoring rubric informed by published neuroscience and behavioural-science literature, combined with AI interpretation of your creative assets.

Our predictions are informed by the same research that guides physical neuroscience labs — we make those findings accessible without requiring R4M+ in lab equipment and 6-8 weeks of study time.

Think of it like a weather forecast: meteorologists don't create weather — they predict it using models trained on historical data. We predict neural engagement using a scoring rubric informed by published neuroscience and behavioural-science literature.

A creative score estimates properties of the asset under a defined assessment method. It does not estimate realised attention, recall, persuasion or conversion in a live placement without explicit environmental, audience, delivery and validation evidence.

Aggregation (DEC-K-001): self-serve scores are a single run (k=1). Concierge scoring uses the arithmetic mean of three independent runs (mean-of-3) — never the median — so the published repeat-scoring SEM applies without a separate calibration study. Uncertainty is shown as a repeat-scoring band (±1.96 × SEM / √k), not as a generic "confidence: N%". Constructs whose k=1 band exceeds δ_abs=5 are released as comparative-only (ordering), not as absolute point estimates.

Every analysis runs through a three-stage pipeline:

Campaign Intelligence scores 2–20 assets in parallel, then synthesizes cross-asset patterns on Claude Sonnet 4.6.

• Per-asset scoring — each file runs the full pipeline (literature-grounded dimension scores, 42-principle audit, geographic conditioning).
• Cross-asset synthesis — ranked variants, consistency analysis (visual, tonal, message alignment), and a kill list of underperformers to cut before media spend.
• Deferred Advisor — batch runs stay fast; full advisor depth generates on demand when you open an individual asset report.
• Deliverables — interactive campaign dashboard plus 11-section print-ready PDF exports for client presentations.

Buyology outputs are labeled with evidence tiers T1–T4. Tier labels state what each score is (and is not): pattern identification, dataset-bound classification proxy, or stronger claims only when the underlying validation supports them.

We commit to the incremental-validity study documented at docs/validation/incremental-validity-study-protocol-v1.md. The study tests whether neuroscience-informed proxies add predictive value beyond cheaper saliency and heuristic baselines. A public URL will be linked here when the protocol is published.

Study activation triggers: a research grant, paid client sponsorship, or a university partnership that funds panel infrastructure. Until then the protocol stands as a public commitment, not a completed claim.

If the study finds null or negative incremental validity, we will publish those results honestly — including where a simpler proxy is sufficient.

No predictive model is perfect. Here's what you should know:

• Ordering is the strongest claim. NeuroScore reliably orders creatives that differ by 2 points or more (median rank agreement 0.92), measured on SA public retail creative (n=20). Below a 2-point gap, treat the difference as not established.
• Demographic calibration is inference-based, not empirically measured per segment. We don't have separate neural recordings for each LSM tier or province.
• Our attention heatmaps use computational saliency models, not actual eye-tracking hardware. They predict where attention is likely to land based on visual properties.
• Individual variation is real. Our models predict population-level tendencies, not how any single person will respond.
• Intention (not live): we intend to calibrate against customer outcome data when a licence-clear outcome corpus and feedback mechanism exist. Neither exists today — no customer outcomes update the live scoring path.

Planned validation work:

• Planned: reliability studies on a frozen licence-clear regression set (S1). The set is currently empty (T8: 0 eligible assets) — S1 has not run. Blocker: no licence-clear members in the frozen evaluation pool.
• Planned: customer outcome tracking to measure prediction-to-performance accuracy (no live feedback loop yet).