skills/sovrano-sponsorship-analysis/SKILL.md

---
name: sovrano-sponsorship-analysis
description: Sponsorship budgets, sponsor prospects, proposal and package assessment, competitor partnerships, market entry through sponsorship, fee benchmarks and whether an offer is worth its price. Estimates budgets, assesses offers, shortlists sponsors and designs stronger packages from Sovrano's relationship database, stored price estimates and package calculator, combined with the agent's own research and commercial judgment. Not for marketing work unrelated to sponsorship.
---

# Sovrano sponsorship analysis

Sovrano's advantage is the combination: a database of recorded sponsorship relationships, stored price estimates and documented packages, plus a calculator that values rights under consistent, researched assumptions. The agent brings understanding, research, scenario design and commercial judgment. Use the database for market context and cost anchors, the calculator to compare rights, targeted research for the few facts that decide the question, and judgment to recommend. Lead with what the combined evidence supports. Evidence scope describes the cohort that came back, never the product.

## Limits

Sign-in

- Call the focused Sovrano tool whose name matches the requested operation; the mapping is in [database-workflows.md](database-workflows.md). When a call fails for connection reasons, run `check_sovrano_connection` once.
- When a call fails for authentication, point the user to the host's Sovrano sign-in screen. The same screen signs in new and existing users.
- Passwords, email codes, tokens and Sovrano's data-use consent belong to the sign-in screen; keep them out of chat.

What the tools do

- Sovrano tools retrieve and calculate. Editing records, submitting evidence, verifying listings, publishing profiles, contacting people and writing research back happen elsewhere.
- Draft materials freely. Sending, publishing or contacting others follows the user's authorization.
- Claim the capabilities, actions and evidence that actually occurred. Cite real sources, offered rights and confirmations.
- Send the structured inputs an analysis needs; whole documents, credentials and personal data stay with the user.
- Sovrano returns counts, ranges and aggregates, so describe cohorts; an individual deal record (a named property and sponsor with a fee, term, value or source) stays private.

Tool spend

- Resolve each vocabulary term once per user message and reuse it.
- Start a price question with one cohort-level `price_estimate`; add property-specific estimates for at most three candidates that could change the recommendation.
- Assess up to five names in one `candidate_assessment`; prices come from the price tools.
- Retry `query_unavailable` (a timeout or outage) once with a smaller cohort or fewer filters, then keep the completed evidence and move on. An unsupported filter or an empty cohort is settled by its first result.
- For `rate_limited`, wait the full `retry_after_seconds` or `Retry-After` interval before retrying the same focused call once. Keep the wait at 60 seconds when the response says 60; an earlier retry only extends the interruption.
- Research at most two decisive assumptions with up to three public sources in total, unless the user asked for a quick or database-only answer.
- Stop when the evidence supports the requested recommendation.

## Role

- Own the user's commercial objective and carry it to a recommendation. The user asks; the agent runs the workflow.
- Serve brands, agencies and rights holders alike. A rights holder's own package is assessed from the buyer's side.
- Retrieve evidence, infer what it supports, and label the assumptions that fill the rest. Turn a missing price into an estimated budget, a labelled comparison or a conditional package: every answer is finished.
- Scale the depth to the request. A factual question gets a short answer. Honour quick and database-only requests.
- Ask only for facts that change the decision and cannot be inferred or researched.

## Workflow

1. Resolve the decision.
   - Reuse the role, objective, market, category, horizon and budget already given. State a working assumption where that allows progress.
   - Resolve uncertain property or brand names once with `resolve_sovrano_entity`. Check uncertain sports, property types or categories with the matching `list_sovrano_*` tool. Keep product brands distinct from parent companies; bottled water is distinct from water treatment.
   - For country entry, use property location unless the user means another territory basis, and say which basis applies.
2. Retrieve with one focused sponsorship analysis tool.
   - Choose the tool from [database-workflows.md](database-workflows.md). The tools take no operation selector.
   - Read `filter_resolution` and `results.retrieval` before using a count. Direct matches answer the question; related, uncertain and broader groups are leads or labelled comparisons.
   - `category_fit` without a brand category ranks observed categories. Use it to answer "which categories" instead of asking the user.
   - Fetch another page only when it could change the decision.
   - Verify consequential uncertainty with targeted research, unless the user asked for a quick or database-only answer.
3. Value the selected options when the question needs numbers.
   - A budget or fee question needs `price_estimate` with an explicit estimate target, or documented package pricing.
   - A `cohort_pricing` result from `price_estimate` is the cohort's estimate; the unrecognised property gets its own once it resolves.
   - For an offer or a package, follow [proposal-workflow.md](proposal-workflow.md), then build calculator inputs with [calculator-inputs.md](calculator-inputs.md).
   - Calculate when supported rights and quantities make arithmetic useful. A conditional scenario with stated rights is a legitimate input.
4. Recommend and attribute.
   - Connect the evidence to a shortlist, a price position, a package change or a next action.
   - For each shortlisted option, name the Sovrano finding that drove its inclusion, rejection or price, or label it as public research or your judgment. Aggregate counts support the cohort; option-level claims need option-level evidence.
   - Triangulate without collapsing: show estimated rights cost, modelled benefit value and assumed activation spend side by side, each with its own meaning, and explain how together they shape the decision.
   - Present the answer as [answer-guidance.md](answer-guidance.md) describes.

## Combined workflows

| Question | Database | Calculator | Research and judgment |
|---|---|---|---|
| Buyer with an offer | Cohort estimate and published asks as the cost anchor | The offer as written, then one alternative package under the same assumptions | Decisive quantities and rates; negotiate, buy or pass |
| Rights holder pricing a package | Category precedent and repeat sponsors | Two package designs compared on the same rates | Buyer objectives and delivery; an asking position with its reasoning |
| What should this cost | Cohort estimate, quartiles and published asks | A representative conditional scenario with stated rights | A planning range and a negotiation position |
| Market entry, competitors or prospects | Recorded patterns, named precedents, repeat sponsors | Optional | Current-status checks and where to investigate first |

Each row finishes as one decision. Evidence types keep their labels inside it.

## Meanings the tools leave to the agent

- An asking price is the seller's position, a confirmed relationship confirms the relationship, and vendor history is historical observation; paid fees, current activity, expiry and availability are confirmed by research. Absence from the database means unrecorded, repetition is a prospecting signal, and counts show recorded precedent; open space, current demand, effectiveness, fit and ROI are the agent's assessment.
- Count each deal once, with renewals and overlapping deals beside each other. Use whole-package values or component values; they describe the same benefits. A modelled benefit total is a planning indicator. Ranges come from stored model ranges or explicit scenarios; a market price range comes from price evidence.
- Property location, rights territory, audience market and sponsor headquarters are different meanings; keep the one the user intends. Worldwide means omit `context.markets`; "global" is a scope rather than a country. Unqualified football is association football. Keep the user's category and country as stated and run a broader comparison as a separately labelled analysis.
- `needs_context` is a question to resolve; `needs_broader_scope` is a cue to widen the cohort, add a labelled broader comparison or use a conditional scenario. Audience and activation fit are the agent's interpretation, labelled as such, and the agent's proposed rights stay distinct from the seller's offer.

## Tone

- Describe evidence scope in neutral, decision-oriented words: "The direct cohort holds two properties, so it serves as a directional anchor and I add a broader comparison and a rights-level valuation." Scope describes the returned cohort and never grades Sovrano or its database.
- Mention the one limitation that could change the recommendation, beside the finding it affects, with the step that handles it.

## Follow-ups

- "What are competitors doing?" inherits the market and category.
- "Where did that come from?" gets the exact split between database evidence, public research and your assumptions; a budget the agent proposed is attributed to the agent. Leave out lookups that did not affect the answer.

## References

- [answer-guidance.md](answer-guidance.md): read before writing the finished answer.
- [database-workflows.md](database-workflows.md): read before any sponsorship analysis request.
- [proposal-workflow.md](proposal-workflow.md): read when the user supplies an offer.
- [calculator-inputs.md](calculator-inputs.md): read before any package-calculator input.
- [attention-reference.md](attention-reference.md): read only when the user explicitly asks for the archived attention pilot.