Hiring Intent Account Feed

Convert job postings into ranked B2B outbound accounts with explainable hiring-intent signals.

Why Teams Use This

• Prioritizes accounts with active hiring momentum.
• Works with both Greenhouse boards and existing Apify datasets.
• Produces explainable scoring with matched categories and keywords.
• Tracks new-vs-seen changes so repeated runs surface fresh opportunities.
• Ships run-level ROI counters in SUMMARY for stakeholder reporting.

Quick Start (2 Minutes)

1. Run with default input (already prefilled with greenhouseBoardTokens: ["webflow"]).
2. Open dataset view overview for ranked accounts.
3. Check key-value store record SUMMARY for run counters and ROI metrics.

What It Does

• Ingests jobs from one of two sources:
  • Apify dataset (sourceDatasetId)
  • Greenhouse Job Board API (greenhouseBoardTokens)
• Normalizes and deduplicates jobs.
• Aggregates signals by company domain.
• Scores accounts using configurable role rules.
• Tracks new-vs-seen signals across runs using persisted state.
• Outputs ranked accounts to the default dataset.
• Writes run metrics and ROI metrics to SUMMARY in default key-value store.
• Optionally posts digest and failure alerts to webhooks.

Best-Fit Use Cases

• Outbound prospecting queues for SDR/AE teams.
• Weekly account prioritization for RevOps.
• Agency lead-list enrichment workflows.
• Signal-driven routing into CRM or outreach automation.

Input Contract

Use the Actor input schema in the Apify run UI.

Source Selection

Provide exactly one source mode.

Field	Type	Required	Notes
sourceMode	string	Yes	greenhouse_api or dataset. Runtime validation enforces consistency with source fields.
sourceDatasetId	string	Yes (if greenhouseBoardTokens is empty)	Apify dataset ID with source job rows.
greenhouseBoardTokens	string[]	Yes (if sourceDatasetId is empty)	Greenhouse board tokens or board URLs. Example: webflow or https://boards.greenhouse.io/webflow.
greenhouseIncludeContent	boolean	No	Greenhouse mode only. Default true. Includes full job description content when enabled.

Source selection is strictly XOR-validated at runtime (schema requires explicit sourceMode and mode guidance):

• dataset mode: non-empty sourceDatasetId and empty greenhouseBoardTokens
• greenhouse mode: non-empty greenhouseBoardTokens and empty sourceDatasetId

Default public run behavior:

• The input UI is prefilled with greenhouseBoardTokens: ["webflow"] so first-time runs work without extra setup.
• To use dataset mode, clear greenhouseBoardTokens and set sourceDatasetId.

Processing Controls

Field	Type	Default	Notes
maxSourceItems	integer	5000	Max source jobs to read in a run (hard-capped at 20000).
maxAccounts	integer	200	Max ranked accounts to output (hard-capped at 1000).
minAccountScore	integer	25	Accounts below this score are dropped.

Dataset Mapping (dataset mode)

fieldMapping supports custom source key lookup for:

• companyNameFields
• companyDomainFields
• companyUrlFields
• jobTitleFields
• jobUrlFields
• locationFields
• postedAtFields
• descriptionFields

Scoring Rules

roleRules[] entries:

• name
• weight
• keywords[]
• negativeKeywords[] (optional)
• outreachAngles[]

Delivery and Billing

Field	Type	Default	Notes
slackWebhookUrl	string	""	Optional digest webhook.
digestTopN	integer	10	Number of rows in digest.
alertWebhookUrl	string	""	Optional failure alert webhook.

Billing note: pay-per-event charging uses a fixed internal event name (ranked-account) and is not configurable via Actor input.

Input Examples

Dataset Mode

{
  "sourceMode": "dataset",
  "sourceDatasetId": "abc123XYZ",
  "maxSourceItems": 3000,
  "maxAccounts": 150,
  "minAccountScore": 25,
  "fieldMapping": {
    "companyNameFields": ["company", "companyName"],
    "jobTitleFields": ["title", "jobTitle"],
    "jobUrlFields": ["url", "jobUrl"]
  },
  "slackWebhookUrl": "",
  "digestTopN": 10,
  "alertWebhookUrl": ""
}

Greenhouse Mode

{
  "sourceMode": "greenhouse_api",
  "greenhouseBoardTokens": [
    "webflow",
    "airtable",
    "https://boards.greenhouse.io/discord"
  ],
  "greenhouseIncludeContent": true,
  "maxSourceItems": 3000,
  "maxAccounts": 150,
  "minAccountScore": 25
}

Source Row Shape (dataset mode)

A typical source row:

{
  "companyName": "Acme Inc",
  "companyUrl": "https://acme.com",
  "jobTitle": "Senior RevOps Manager",
  "jobUrl": "https://jobs.acme.com/revops-123",
  "location": "San Francisco, CA",
  "postedAt": "2026-02-01T12:00:00Z",
  "description": "Own Salesforce administration and RevOps process improvements."
}

Output Contract

Default Dataset

One row per account.

Field	Type	Description
accountKey	string	Stable account identifier used for state/dedupe/ranking.
companyName	string	Account display name.
companyDomain	string | null	Canonical website domain when available; null if unresolved.
score	integer	Final account score (0-999).
jobCount	integer	Total jobs after normalization/dedupe.
matchedJobCount	integer	Jobs that matched at least one rule.
matchedJobRatio	number	matchedJobCount / jobCount.
newJobCount	integer	New jobs vs persisted seen-state.
newMatchedJobCount	integer	New matched jobs vs seen-state.
isNewCompany	boolean	Whether account was unseen in prior state.
matchedCategories	string[]	Rule categories with hits, ordered by hit strength.
outreachAngles	string[]	Up to 4 angle suggestions from matched categories.
latestPostedAt	string	Latest posting timestamp observed for the account (ISO 8601 UTC).
avgSignalConfidence	number	Mean confidence across matched rule events.
recencyScore	integer	Recency factor (1-100).
scoreBreakdown	object	Components used to build final score.
signals	object[]	Up to 8 matched signals.
generatedAtUtc	string	Run timestamp in UTC (Z).

signals[] fields:

• title
• location
• postedAt
• jobUrl
• matchedRules
• signalConfidence
• matchedKeywordsByRule

scoreBreakdown fields:

• categoryScore
• volumeComponent
• qualityComponent
• confidenceComponent
• recencyComponent

SUMMARY (default key-value store)

SUMMARY includes run counters plus ROI summary.

Field	Type	Description
sourceType	string	dataset or greenhouse_api.
sourceDatasetId	string	Dataset ID (empty in Greenhouse mode).
sourceItemsRead	integer	Jobs read from source ingestion.
accountsMatched	integer	Accounts with non-zero matched score before cap.
accountsOutput	integer	Accounts actually delivered to dataset (after score/output caps and PPE cost cap, if hit).
generatedAtUtc	string	Run timestamp UTC.
filteredMissingDomain	integer	Jobs dropped for missing domain.
filteredMissingTitle	integer	Jobs dropped for missing title.
filteredDuplicateJobs	integer	Duplicate job fingerprints dropped.
filteredBelowMinScore	integer	Accounts dropped below minAccountScore.
roiSummary	object	Aggregate output quality metrics.
stoppedByMaxTotalCharge	boolean	True when PPE max total charge limit stopped output early.
delivery	object	Delivery accounting for attempted vs delivered vs charged rows.
ingestion	object	Present in Greenhouse mode; board-level ingestion stats.

roiSummary fields:

• outputMatchedJobs
• outputNewMatchedJobs
• outputHighConfidenceAccounts
• highConfidenceThreshold
• outputAvgAccountScore
• outputAvgMatchedJobRatio
• outputAvgSignalConfidence

delivery fields:

• attemptedRows
• deliveredRows
• chargedRows
• chargedEventName
• pricingModel
• isPayPerEvent
• knownPayPerEvents
• eventChargeLimitReached
• stoppedByMaxTotalCharge

ingestion includes board-level success/failure detail:

• succeededBoards[]: { token, jobsFetched }
• failedBoards[]: { token, reason, httpStatus?, message }

Processing Behavior

Normalization and Dedupe

• Domains are normalized from explicit domain, company URL, or job URL.
• Known aggregator hosts are re-keyed to account-level keys (for example webflow.greenhouse).
• Description text is cleaned (HTML/boilerplate removal) before scoring.
• Dedupe uses a deterministic fingerprint over domain/title/location/url (or description snippet fallback).

Scoring

• Rule hits come from keyword matches in title + cleaned description.
• A rule match requires either a title hit or at least two distinct keyword hits.
• Optional negativeKeywords suppress matches for known noisy contexts.
• Rule-level confidence is computed from title/body hit mix and keyword coverage.
• Final score blends:
  • weighted category hits
  • match ratio quality
  • confidence quality
  • recency
• Accounts with no matched rules are not emitted.

State and Newness

• State stores seen fingerprints per accountKey.
• newJobCount and newMatchedJobCount are calculated against persisted state.
• Seen-state is updated for all matched domains, not only capped output rows.

Failure Handling

• Failures are logged with pipeline stage context.
• If alertWebhookUrl is set, a failure alert message is posted.
• In Greenhouse mode, per-board failures are counted in ingestion.greenhouseBoardsFailed; successful boards still process.

Changelog

See CHANGELOG.md.