skills/sovrano-sponsorship-analysis/attention-reference.md

# Archived attention scenarios

Use `model_sponsorship_attention` only when the user explicitly asks to inspect the archived, uncalibrated pilot. The pilot covers fifty properties and returns three alternative allocation scenarios from I = R x E x M, then I x A. R is researched attendance, capacity with an assumed occupancy, or an explicit 1,000-visit normalization. E and A are policy assumptions. M uses historical consumer-admissions monetization priors with geography pooling and broad proxies. The model is not calibrated to sponsorship transactions.

Before interpreting any amount, read `estimate_type`, `inputs.reach_kind`, `period`, `edition`, `range_meaning`, `monetization_prior`, `transaction_calibration`, the sources, the sensitivity and the unvalued components.

- A normalized result is per 1,000 visits. The property's total value and attendance are separate inputs.
- Capacity is not attendance. Keep one-match, full-event and annual scopes separate.
- The three shares (0.1%, 0.3%, 1%) are alternative uncalibrated allocation assumptions; packages, confidence levels and market price points are separate concepts. Each share stands alone.
- The 10th to 90th percentile ranges describe assumptions; calibrated prediction intervals would need transaction data.
- Named sponsor fees, current offers, inventory entitlements and availability come from other evidence.
- Currency conversion is dated and separate from the historical nominal values; today's prices come from current evidence.

Example request: `{"subject":{"name":"Manchester City"},"proposal":{"currency":"CHF"},"options":{"max_examples":1}}`.

Presentation: lead with "conditional onsite budget scenario" and state whether R comes from reported visits, capacity, sold passes or a 1,000-visit normalization. State the selected A share and the edition or period beside the number. Disclose that A has not been calibrated against sponsorship transactions. For normalized results the headline says "per 1,000 visits" and names no total property budget. Explain the most influential uncertainty and leave unavailable rights unvalued.

When an attention-model property also has published package prices, call `find_sponsorship_asking_prices` before recommending a budget. A documented ask for the actual offered package is the offered cost; it is distinct from the estimated deal cost and the benefit valuation. A difference can reveal a different prior or different rights; treat it as a scope difference until package scope, edition and consideration are confirmed to match.