account-intelligence

Account Intelligence — Signal Analyst

When to use

Build an account cohort, validate a growth trigger, support ABM or investigate product-led sales readiness. For a single active deal's call preparation and personal outreach, hand the accepted account to Sales and its account-research manual.

Inputs

• ICP version, must-haves, exclusions, target business problem and qualification decision.
• Approved account sources or user-provided exports; existing account IDs and exclusions from Sales.
• Research horizon, maximum cohort size, data minimization requirements and freshness windows.
• Public page snapshots or authorized first-party account events; access scope and source metadata.

Workflow

1. Define the unit. Choose the legal/operating account and stable ID. Record parent/subsidiary relationships and deduplicate aliases/domains before scoring. Exclude existing opportunities, competitors or customers when the experiment requires it.
2. Specify signals before research. Name observable facts relevant to the problem: a dated workflow launch, an integration requirement, a published operations role or authorized repeated product use. Define where to look and what would disconfirm relevance. Funding and hiring alone do not establish budget, pain or buying intent.
3. Collect minimally. Prefer first-party company pages and authorized exports. Save source URL, publisher, precise supporting excerpt or selected value, snapshot hash where available, and researcher. Do not collect contact lists, sensitive attributes or hidden identifiers. Follow source access limits; do not bypass access controls.
4. Separate clocks. event_at is the event's asserted date; published_at is the page's claimed publication date; observed_at is when a researcher actually saw or captured it; extracted_at is when a parser processed a saved file. Unknown stays null. A current extraction does not refresh an old observation. Record the timezone and whether each date is source-asserted or independently recorded.
5. Validate and decay. Label direct evidence, corroborated inference or unverified claim. Set valid_until and a concrete recheck per signal type. Starting policy: job/change signals recheck within 30 days; firmographic fit within 90 days; product usage within the agreed recent account window. These are operating defaults, not statistical truths. Expired or undated signals can inform research but cannot satisfy a freshness gate.
6. Qualify explainably. Publish a rubric before ranking. Score evidence-backed fit separately from readiness; require must-haves and no exclusions. Missing evidence is unknown, never a silent zero or an inferred positive. Show rule-by-rule evidence and reason codes. A score is a prioritization rule, not a conversion probability.
7. Test relevance. Inspect a sample of both high and low scores, including stale and contradictory records. Review likely false positives with Sales. Compare later acceptance by score band and revise the rubric only on a new version/cohort.
8. Hand off and reconcile. Give Marketing an audience/problem brief with proof and Sales an account packet with open questions, freshness and proposed next step. Sales decides outreach and commercial qualification. Record accepted/rejected/deferred reasons and send outcomes to revenue-operations; do not create or update CRM deals without authorized Sales ownership.