HG Tool Chain: Account Research

The canonical order of HG calls for researching one account, with the gate at each step and what it costs.

When to use

• Authoring a new account-research workflow or brief.
• Auditing a workflow that calls tools in an unprincipled order.

Tools you'll touch

• company_firmographic: resolve the company and its level. Cost: 0.1 per company (see hg-credit-awareness).
• company_hierarchy: ownership and sister companies. Cost: 0.1 per node.
• company_technographic: the stack. Cost: 2 per company.
• company_spend / company_ai_spend: size of the opportunity. Cost: 3 / 2 per company.
• company_enrich: several sections in one call. Cost: sum of sections per company.
• list_intent_topics then company_intent: what they're researching. Cost: free, then 0.1.
• company_fai: which department to target. Cost: 2.
• contact_search then contact_enrich: named people. Cost: 2, then 0.2 per email.
• company_contracts: outsourcing context. Cost: 1 per company.
• web_search then sec_filing_section / sec_full_text_search: ticker, then filings. Cost: 0.05, then 1.

The chain

0. check available tools (hg-mcp-version-and-availability)
1. company_firmographic            → resolve, check company_level
2. company_hierarchy (optional)    → ownership
3. company_technographic + company_spend   (or company_enrich)
4. list_intent_topics → company_intent
5. company_fai
6. contact_search → contact_enrich
7. company_contracts (optional)
8. ticker via web_search → SEC tools (public companies only)

Each arrow is a gate: don't run the next step until the previous output justifies it.

Why this order

1. company_firmographic first. Every later step needs the hg_id and a deliberate choice of entity. Check company_level: if the domain resolved to a subsidiary and the user meant the brand or group, re-query on corporate_parent_id (brand) or global_hq_id (ultimate parent). Carry the hg_id forward; keep the domain for company_cloud_spend and gov tools.
2. company_hierarchy when ownership matters ("who owns them", sister companies, regional entities). Pass the resolved hg_id, never a guessed domain. Use mode:"parents" for ownership; use depth:5 with filters for subsidiaries, and dedupe nodes by id.
3. company_technographic, then company_spend. The stack gates whether there is anything to say; spend sizes the deal. Resolve product or category ids first and pass them as filters, with install_fields to keep the payload small. Read intensity within the company only. Add company_ai_spend for AI plays.
   • Option: for several sections on one to a few companies, company_enrich with explicit fields (e.g. ["firmographics","technographics","spend"]) does it in one round-trip. Sections are capped per company, so use the dedicated tools when you need the full install list or filters.
4. Intent. Resolve topics with list_intent_topics (one lowercase keyword), then call company_intent with topic_ids, fields:["topics"] and granularity:"global". Without resolved topics the result is saturated and means nothing. No topics at a smaller company means no coverage, not no interest.
5. company_fai before any contact lookup. It gives the department and the decision_maker_titles / influencer_titles to search for. See hg-fai-vs-contact-search.
6. contact_search with the FAI titles, then contact_enrich with revealPhone:false for the few people you'll actually quote.
7. company_contracts for GSI and ICT outsourcing context: which integrators run parts of their IT, and recent announced deals. It is not a renewal tool: no renewal-likelihood or expiry-confidence fields. "Days to end date" is fine without a confidence claim. A missing contract proves nothing.
8. SEC. No HG tool returns a ticker, so resolve it first (web_search, e.g. "Acme Corp investor relations ticker"). US filers: sec_filing_section (10-K, 10-Q, 8-K). Foreign filers (20-F, 6-K): sec_full_text_search with tickers, formTypes, query and a startDate. If no ticker exists, skip this step.

Credit estimate per chain

Using the proportional table, for one company:

Chain	Steps	Credits
Quick profile	1, 3, 4	about 5.2
Standard brief	1, 3, 4, 5, 6 (5 emails)	about 10.2
Full brief	Standard + 2 (about 20 nodes), 7, 8	about 14.3
Add-ons	company_ai_spend +2, company_ai_maturity +2, company_install_time_series +3 per product

Multiply steps 1, 3 and 7 by the number of companies when batching. Cached calls cost 0.

Where to drop a step

• No spend for pure technographic-fit plays (competitive replacement).
• No hierarchy when the user named a specific entity and ownership doesn't matter.
• No contracts for awareness or discovery work.
• No SEC when there's no ticker or the deliverable doesn't need filings.
• No contacts for territory sweeps that only need account-level summaries.

Where to add a step

• Intent before technographic when the primary signal is "what are they researching".
• company_install_time_series with product_ids when the question is adoption trend.
• web_search for recent news or non-US registries; see hg-fallback-to-web-search.

Common pitfalls

1. Calling contact_search before FAI.
2. Calling SEC tools without resolving a ticker, or sec_filing_section for a foreign filer.
3. Using the user's domain without checking company_level.
4. Running intent without topic_ids.
5. Framing contracts as renewal timing.
6. Running the full chain over a territory. Pre-filter with search_companies and firmographic, then run the chain on the top 10 to 20.

Reference

• V2 tool docs: https://phoenix.hginsights.com/docs/mcp-tools/v2/company-enrich
• Full docs for LLMs: https://phoenix.hginsights.com/docs/llms-full.txt
• Related skills: hg-mcp-calling-rules, hg-firmographic, hg-subsidiary-handling, hg-technographic, hg-spend, hg-intent, hg-fai-vs-contact-search, hg-contracts, hg-sec-research, hg-credit-awareness