Skill: HG Contracts
Catch renewal windows before competitors do, with displacement plays grounded in real expiry data.
Overview
Time outreach to live renewal windows and surface displacement opportunities competitors miss. Claude learns the 122-day window where renewal cycles are actually in motion, treats expired-and-not-replaced contracts as whitespace plays, and reads renewal-likelihood as a signal to act on rather than a forecast to quote.
Use cases
Territory sweeps that surface in-cycle deals
Claude flags every contract in the territory expiring inside 122 days, sorted by modeled deal value. The output is a ranked list of decisions-already-in-motion — the conversations where 'have you considered alternatives' lands as a real question, not a fishing trip.
Whitespace plays you'd otherwise miss
When HG sees a contract that expired more than 90 days ago and no replacement on the same product family, the brief flags it as whitespace — the company either churned out of the category or self-built. Worth a discovery call.
View full skill
HG Contracts
When to use
- A workflow needs to time outreach to a renewal window.
- A prompt is identifying displacement opportunities (expired-not-replaced contracts).
- An author is sizing a deal —
company_contractsgives a modeled value to anchor on.
Tools you'll touch
company_contracts— modeled contract data: deal value, expiry, renewal likelihood
What HG actually returns
For a given company, HG returns the contracts it has modeled across publicly observable signals (procurement filings, press releases, partner announcements, hiring data, etc.).
Important framing:
- Modeled deal value, not invoiced. The dollar figure is HG's estimate of annual run-rate or total contract value, depending on what it could observe. Treat as order-of-magnitude.
- Expiry-date confidence. Each contract carries a confidence band on the expiry date — high confidence when the source explicitly stated the term, lower when HG inferred from a typical contract length.
- Renewal-likelihood scoring — a modeled probability that the company renews vs. switches. This is a signal, not a forecast.
The Phoenix-side credit cost is 3 credits per result returned (pricing.ts). For territory-wide sweeps this adds up — paginate aggressively.
How to read it
The 122-day expiry threshold for active displacement. Inside 122 days, the renewal cycle is usually live: the procurement team is gathering RFPs, the incumbent is in a renewal motion, and the buyer is in market. Outside 122 days, an outreach is "planning conversation"; inside, it's a "decision conversation". This threshold is a working heuristic — see hg-insights-api.md#contracts for the underlying confidence model.
Expired-not-replaced as whitespace. When HG sees a contract that expired more than 90 days ago and no new contract has been observed for the same product family, that's a whitespace signal — the company either churned out of the category or self-built. Worth a discovery call.
Renewal-likelihood readings:
-
0.7 → likely renews; lead with expansion or competitive insertion, not displacement.
- 0.4-0.7 → genuinely contested; this is where displacement plays land.
- < 0.4 → already losing the renewal; if they're on the incumbent's product, ride that signal hard.
Common pitfalls
- Quoting the modeled value as if it were a price. "Their contract is $X" reads like fact. Hedge: "HG models the contract at ~$X annual run-rate".
- Ignoring the expiry-date confidence band. A low-confidence date inside 122 days is still meaningful, but treat the 122-day window as approximate.
- Treating renewal-likelihood as a forecast. It's a signal HG fits to public observables. The actual outcome depends on a dozen things HG cannot see.
- Forgetting credits scale with results. A territory query that returns 200 contracts costs 600 credits. Filter aggressively (industry, employee band, expiry window) before opening the floodgate.
Citation rules
Always cite company_contracts results with the modeled qualifier on first use. Format: "HG-modeled contract: $X annual run-rate, expires Mar 2026 (high confidence)". On subsequent mentions, drop the framing.