Use Lead Scoring → Research & Signals to decide which external sources scoring can consult, which research tasks it can request and which customer benchmarks it can compare. You need scoring-policy authority and approval for the intended providers and spending. A configured source does not prove credentials, coverage or provider readiness.
What you’ll see
Configure the research draft
- Expand Signal sources. Review each source’s Label, Source key, TTL days, Enabled, Allowed profiles, and Blocked presets. Use Add source only for a supported provider. Review Requires manual confirmation before spend rather than assuming every enabled source may run automatically.
- Review Risk handling, including Negative news mode and Allow automatic disqualification. These choices can affect treatment of leads and deserve an explicit policy decision.
- Open Research tasks. Use Add task to configure the label, provider, tier, enabled state, Cost tier, Max cents, Cache days, signal types, required inputs and applicable profiles. Check Rescore after completion separately.
- In Benchmark customers, review Benchmark scoring, active set references, Anchor bands, and Similarity inputs. Add set references a benchmark set; it does not create a verified customer dataset from a name alone.
- Choose Save draft. Publish only the reviewed policy, checking the shared Rescore on publish option and all pending changes from other scoring tabs.
Confirm and recover
Confirm the stored policy version and references after reopening. Empty sources, missing tasks or unavailable benchmark data must not be described as successful research. If publishing succeeds but rescore queueing fails, retain that distinction and inspect Operations before retrying. Failed policy loading can leave layout controls visible; restore the connection before editing or drawing conclusions.
Screenshots
The screenshot for this guide is shown above, under “What you’ll see”.