// How it works
From scattered renewal data to one canonical decision
Search broadly enough to protect recall. Resolve renewal timing strictly per Company. Preserve uncertainty. Automate only when timing and Renewal Risk are trustworthy enough.
1. Portal-wide candidate discovery
Renewal Radar builds a deterministic candidate inventory from safety cohorts, existing Renewal Radar state, and eligible renewal evidence across Company, Deal, Subscription, Line item, Quote and relevant custom-object sources.
The expensive account-level activity and risk lookups then run on that cohort. If HubSpot pagination, search caps, permissions or association coverage make complete recall uncertain, Renewal Radar fails open to the broader Company inventory rather than silently skipping possible renewals.
2. One RenewalResolution per Company
Eligible Company, Deal, Subscription, Line item, Quote and relevant custom-object evidence enters the same canonical resolver. Each Company resolves independently to one timing state:
- Explicit: direct renewal or contract-end timing.
- Proxy: renewal-specific evidence weaker than an explicit date.
- Estimate: derived monitoring timing, such as an eligible closed-won Deal plus the configured contract term.
- None: no reliable current renewal timing was resolved.
3. Supported does not mean every date
Broad source support stays strict. Subscription timing must represent genuine end or renewal timing. Line-item timing must pass the required Deal route and commercial context. Quote timing must pass the contract/renewal and accepted/signed rules. Custom objects need eligible contract/renewal context plus a valid Company or Deal association route.
Generic Quote expiration, billing/payment cadence dates, forecast or operational dates, created/modified timestamps and ambiguous generic dates remain excluded.
4. Timing confidence is separate from risk severity
High / Medium / Low timing confidence answers “how trustworthy is the renewal timing?” Renewal Risk severity answers “how urgent is the account?” They are separate by design.
Estimates remain Low-confidence and task-ineligible. Material disagreement between independent explicit sources, or incomplete authoritative timing, also blocks task eligibility.
5. Renewal Risk consumes the canonical date
Renewal Risk evaluates the account around the canonical renewal timing. Renewal-specific motion, such as a future meeting before renewal or an active renewal Deal, can affect urgency. Generic calls or replies, raw ticket count and other activity stay supporting context; they do not by themselves lower or raise Renewal Risk.
The configurable renewal alert window is 30, 60, or 90 days. A renewal inside the final 14 days remains High urgency even when renewal motion exists; very near renewal with no renewal-specific motion can be Urgent. High urgency does not bypass the timing-trust gate.
6. Engagement Risk is supporting evidence
Engagement Risk uses deterministic metadata such as inbound reply timestamps, meeting timestamps, contact activity, Company hs_last_logged_call_date, and ticket status/counts.
It stays a secondary Activity Check: useful context, but it does not outrank or replace the primary Renewal Risk decision and does not create a standalone automatic Renewal Radar task.
7. The production task gate
A new automatic renewal task requires task-eligible canonical timing, Company task eligibility, actionable Renewal Risk, and permission from the existing task lifecycle and account controls.
Medium/Low explicit timing, proxies, estimates, no timing, material conflicts, incomplete authoritative timing, and Engagement Risk by itself cannot create a new task.
8. The decision layer for HubSpot agents
Renewal Radar’s four approved Agent Tools let HubSpot agents retrieve at-risk accounts, explain company risk, snooze an alert for 14 days and record an explicitly confirmed renewed or churned outcome.
The two read tools consume existing decisions. The action tools use the same risk-control services as the app. Recording an outcome completes the open follow-up task; snoozing changes app state without changing CRM properties. Agents must not infer outcomes from account activity.
Agent Tools require active Team or Partner monitoring and must be added to your HubSpot agent. Follow the setup guide. The decision engine uses deterministic rules; the agent’s wording can vary.
Renewal Radar is powered by Sighub, the company building decision infrastructure for agents and humans.
9. Large-portal validation
The latest production read-only validation considered 24,159 Companies, processed 375 canonical candidates and completed in 12m33s with zero failed Companies, zero source-positive candidate misses and zero legacy safety misses.
CRM Search ran at 3.92 requests/second: 2,942 Search requests, zero HTTP 429s, zero retries and zero reactive backoff. Broad-source inventory added one Search request and 244ms in that validation portal. Read the full speed and accuracy benchmark.
