HG Contracts

Read a company's publicly announced ICT outsourcing and GSI deals, and optionally its U.S. federal awards, without overclaiming what the data can prove.

When to use

• A brief needs a company's outsourcing or systems-integrator footprint: which GSIs it works with, for what service lines, at what announced value.
• A displacement or partner play needs to know which contracts are still running and when they're due to end.
• A government-facing account needs its U.S. federal award history (federal mode).

Don't use it for the customer's own SaaS renewals, for renewal forecasting, or as proof that a company has no outsourcing relationships.

Tools you'll touch

• company_contracts: announced ICT outsourcing and GSI contracts for up to 25 companies (hg_ids[] or domains[]), plus U.S. federal awards for one company with include_federal_contracts:true. Cost: 1 per company, billed proportionally to the batch (see hg-credit-awareness). Cached 30 days.
• search_federal_contracts, company_gov_relationships: federal awards by criteria, or a company's prime/sub network. Cost: 1 per call, not cached. See hg-gov-contracting.

What the data is

• Scope: publicly announced ICT outsourcing and GSI deals, plus U.S. federal awards when federal mode is on. Nothing else: no software subscriptions, no vendor invoices, no private agreements that were never announced.
• Curated, not comprehensive. Precision is high, recall is low. A contract that comes back is real; a contract that doesn't come back proves nothing. Report a gap as "not disclosed", never as "no contract".
• No renewal or confidence fields. company_contracts returns no renewal-likelihood or expiry-confidence fields; never infer renewal odds. You may compute days from today to end_date, stated as a plain date difference with no confidence attached.
• Tenure and renewal. Tenure is a weak heuristic for renewal timing, never a renewal date. Say so whenever the user asks about renewal.
• Cite it as a publicly announced contract, not as a modeled figure: "Publicly announced contract (HG): $40M, 5-year term, announced 2023-03-14."

Fields you'll read

Per company: contract_count, total_contract_value_amount and total_matching_contracts. The first two describe the returned page only; total_matching_contracts is the full count across pages.

Per contract (contracts[]):

Field	What It Tells You
vendor_name	The GSI or outsourcing counterparty (in federal records: the awarding agency)
title, summary	What the deal covers
deal_value_amount	Announced value in USD; 0 means undisclosed, not free
start_date, end_date, announcement_date	Contract dates; end_date can be missing
contract_term_months	Announced term
contract_event, bid_process	The deal event (e.g. "New") and how it was bid (e.g. "Competitive")
service_lines, pricing_structure, contract_structure	Scope and commercial shape
customer_drivers	The stated business reason for the deal
signing_country, performance_criteria, larger_contract	Context

How to call it

1. Filter to what's live. Results are sorted by deal value, and large companies can return mostly long-expired deals on the first page. For a "current footprint" question pass active_only:true or end_date_after:"<today>".
2. Remember what "active" means. active_only counts contracts with no end_date as active. Say "active or open-ended" when you rely on it.
3. vendor_name needs the full counterparty name ("Acme Consulting Group", not "Acme"). A fragment can match nothing.
4. Other filters: min_deal_value, max_deal_value, start_date_after, start_date_before, end_date_before, end_date_after, limit (default 50, max 100), offset.
5. Batch by one selector type and match results by organization_id or company_domain, not by position.

Federal mode

• include_federal_contracts:true works for one company per call and needs the SAM.gov integration. If the call fails with a missing-integration message, say federal data isn't available on this plan.
• Always pass company_name with the company's exact legal name alongside the domain.
• Federal awards cover roughly the last 5 fiscal years. Older awards won't appear.
• In federal records, vendor_name is the awarding agency, not a supplier. Read agency detail from federal_data.
• Before saying "no federal awards", check federal_data_status: resolved, contracts_found and errors. A non-empty errors array means the lookup didn't complete; report "federal data unavailable for this run", not "none".
• sam_entity adds registration detail (UEI, CAGE code, business types) when the entity resolves.

Reading it for GTM

• Timing: "Announced 5-year term ending 2027-06-30, about 270 days from today." No probability, no "likely to renew".
• Displacement or partner angle: which GSIs hold which service lines, and whether the deal was competitive (bid_process).
• Sizing: quote deal_value_amount as the announced value. Skip $0 values or say "value not disclosed".
• Absence: "No announced outsourcing contracts in HG's curated set" is the most you can say. Don't call it whitespace.

Common pitfalls

1. Inventing renewal odds or confidence bands. The fields don't exist. Days to end_date is the only timing figure you can give.
2. Reading a missing contract as evidence. Recall is low; "not disclosed" is the correct label.
3. Reading the first page as the current state. It's sorted by value and often dominated by old, large deals. Filter with active_only or end_date_after.
4. Treating $0 as a real value or summing it into totals.
5. Using page totals as company totals. contract_count and total_contract_value_amount cover the returned page; use total_matching_contracts for the count.
6. Reading a federal vendor_name as a supplier. It's the awarding agency.
7. Saying "no federal awards" without checking federal_data_status.errors, or without passing company_name.
8. Using it for the customer's own SaaS renewals. This tool doesn't cover them.
9. Calling the value "modeled". These are publicly announced deals; cite them that way.

Reference

• V2 tool docs: https://phoenix.hginsights.com/docs/mcp-tools/v2/company-contracts
• Full docs for LLMs: https://phoenix.hginsights.com/docs/llms-full.txt
• Related skills: hg-mcp-calling-rules, hg-gov-contracting, hg-citation-discipline, hg-modeled-vs-observed