Schema

Every signal in the Autobound Signal Database follows a normalized schema. This structure applies across all 29 signal sources (SEC filings, news, hiring data, social media, and more), enabling consistent parsing and entity matching.

Quick Navigation

Match signals to your contact/company database → Resolution (join on domain, email, or linkedin_url)

Download data from GCS → Delivery (timestamped folders, JSONL/Parquet)

Browse all signal types → Signal Catalog

Base Schema

All signals share this structure:

{
  "signal_id": "7dfdb4b4-c0b4-4620-aca6-e7263123028e",
  "batch_id": "2025-11-14-09-30-00",
  "signal_type": "job-change",
  "signal_subtype": "jobChange",
  "detected_at": "2025-11-14T00:00:00Z",
  "association": "contact",
  
  "contact": { // use email or LinkedIn url to match. Only shown for contact signals.
    "first_name": "Lashawna",
    "last_name": "Bean",
    "name": "Lashawna Bean",
    "email": "[email protected]",
    "job_title": "Director Of People & Public Affairs",
    "linkedin_url": "https://www.linkedin.com/in/lashawna-bean",
  },
  
  "company": { // use domain or LinkedIn url to match. Shown for both contact & company signals.
    "name": "Harlem Stage",
    "domain": "harlemstage.org",
    "linkedin_url": "https://www.linkedin.com/company/harlem-stage",
    "industries": ["Performing Arts"],
    "employee_count_low": 11,
    "employee_count_high": 50,
    "description": "Harlem Stage ignites the artistic freedom of performing artists..."
  },
  
  "data": { // this is the signal-specific data. Different schema for each signal type.
    "summary": "Joined Harlem Stage as Director of People & Public Affairs 2 months ago.",
    "job_change_occurred": true,
    "promotion_occurred": false,
    "months_since_job_change": 2
  }
}

Note: The association field indicates whether this is a "contact" or "company" level signal. For company-level signals, the contact field will be null or omitted.

Two Types of Signals

Type	Match By	Examples
Company signals	`company.domain` or `company.linkedin_url`	10-K, Reddit, News, Hiring Velocity
Contact signals	`contact.email` or `contact.linkedin_url`	LinkedIn Posts, Job Change, Work Milestone

The association field tells you which type: "company" or "contact".

→ Full matching guide with SQL examples: Resolution

Field Definitions

Core Signal Fields

Field	Type	Required	Description
`signal_id`	string (UUID)	✓	Globally unique identifier for this signal instance. Use this as your primary key and for deduplication - no two signals will ever share the same `signal_id`.
`batch_id`	string		Processing batch identifier (e.g., `2025-11-14-09-30-00`). Signals in the same batch were processed together. Useful for tracking data lineage. Only exists for some signal types where multiple signals are extracted from the same document, like 10k.
`signal_type`	string	✓	The broad category of signal (e.g., `10k`, `contact-linkedin-post`, `hiring-velocity`, `job-change`). Each signal_type is delivered to it's own bucket for most customers.
`signal_subtype`	string	✓	The most important field for filtering. Specifies exactly what was detected (e.g., `ceoChange`, `acquisitionAnnounced`, `aiInvestment`). A single `signal_type` can have dozens of subtypes - this is how you target specific business events.
`association`	string	✓	Whether this signal is about a `"contact"` (person) or `"company"`. Determines which entity object contains the primary data.
`detected_at`	string (ISO 8601)	✓	When Autobound detected this signal. For event-based signals (news, SEC filings), the actual event may have occurred earlier - check signal-specific date fields in the `data` object.

Why `signal_subtype` Matters

The signal_subtype field is how you filter for actionable intelligence. For example:

signal_type	signal_subtype	What it means
`10k`	`ceoChange`	CEO transition disclosed in annual filing
`10k`	`aiInvestment`	AI/ML investment mentioned in annual filing
`10k`	`acquisitionAnnounced`	M&A activity disclosed
`news`	`funding`	Company raised funding
`news`	`hires`	Executive hire announced
`hiring-velocity`	`hiringVelocity`	Overall hiring pace snapshot
`hiring-trends`	`hiringEngineeringRoles`	Engineering department hiring activity

Each signal source has its own set of subtypes. See the Signal Catalog for the complete list.

Contact Object

Contact information for the person associated with this signal. Present for contact-level signals (association: "contact"), null or omitted for company-level signals.

Field	Type	Suggested Matching Key	Description
`first_name`	string		First name
`last_name`	string		Last name
`name`	string		Full display name
`email`	string	Yes ✅	Professional email address. Primary matching key - use this to join contact signals to your CRM or database.
`job_title`	string		Current job title at time of signal detection
`linkedin_url`	string	Yes ✅	Full LinkedIn profile URL, e.g. `https://www.linkedin.com/in/lashawna-bean`. Secondary matching key - use when email is unavailable or for additional validation.

Company Object

Firmographic data for the company associated with this signal. Present on all signals (both contact and company level).

Field	Type	Suggested Matching Key	Description
`name`	string		Company display name
`domain`	string	Yes ✅	Primary website domain (e.g., `harlemstage.org`). Primary matching key - use this to join company signals to your database. Normalized without `www.` prefix.
`linkedin_url`	string	Yes ✅	Full LinkedIn company page URL, e.g. `https://www.linkedin.com/company/harlem-stage`. Secondary matching key - use when domain matching fails or for additional validation.
`industries`	array[string] or string		Industry classification(s). May be a single string or array depending on signal source.
`employee_count_low`	integer		Lower bound of employee count range
`employee_count_high`	integer		Upper bound of employee count range. When `low` equals `high`, this is an exact count.
`description`	string		Company description

→ SQL examples and join logic: Resolution

The `data` Object: What Makes Each Signal Unique

The fields above (signal_id, company, contact) are consistent across all signals. The data object is where each signal type has its unique payload - the actual intelligence detected.

Fields vary by signal_type, but these common fields appear across most signals:

Field	Type	Required	Description
`summary`	string	✓	One-line summary of what was detected. Human-readable, suitable for display in UIs or email subject lines.
`detail`	string		Longer explanation of why this signal matters and what action it suggests. 2-3 sentences typically.
`relevance`	float		Business relevance score from 0.0 to 1.0. Higher = more likely to indicate buying intent or sales opportunity.
`confidence`	string		How confident we are in the signal accuracy: `high`, `medium`, or `low`. Filter to `high` for automated workflows.
`sentiment`	string		Overall sentiment: `positive`, `neutral`, or `negative`. Useful for prioritizing outreach tone.
`source_url`	string		Link to the original source (SEC filing, news article, etc.). Use for verification or to include in outreach.
`competitors_mentioned`	array[string]		Competitor companies mentioned in the source. Useful for competitive displacement plays.
`technologies_mentioned`	array[string]		Technologies, products, or platforms mentioned. Useful for tech stack targeting.

Signal-specific fields vary by type:

Signal	Unique Fields in `data`	Full Schema
10-K Filings	`filing_year`, `fiscal_period`, `excerpts`	View →
News	`published_at`, `article_sentence`, `location_data`	View →
Job Change	`months_since_job_change`, `promotion_occurred`	View →
Hiring Velocity	`numberOfOpenRoles`, `historicalComparison`	View →

→ All signal schemas: Signal Catalog

Podcast

Podcast intelligence ships as two signal types that share one data vocabulary. Both follow the standard envelope above - only the data payload differs.

signal_type	association	Match By
`podcast-contact`	`"contact"`	`contact.email` or `contact.linkedin_url`
`podcast-company`	`"company"`	`company.domain` or `company.linkedin_url`

The data payload takes one of three shapes, determined by signal_subtype:

Family	signal_subtype	What it is
Appearance	`podcastAppearance` (contact), `companyPodcastAppearance` (company)	One row per entity per episode: who appeared, talking points, a company snapshot
Claim	65+ subtypes shared with the SEC/earnings vocabulary: `aiInvestment`, `techAdoption`, `painPointDisclosed`, `strategicInitiative`, `marketExpansion`, `careerHistory`, `fundingEvent`, `executiveOpinion`, `industryPrediction`, ...	One row per distinct on-air business claim, backed by verbatim transcript quotes
Sponsorship	`podcastSponsorship`	One row per sponsor ad read, with the verbatim offer text

A single episode typically yields one appearance row plus several claim rows - join them on episode_id.

Episode Fields (all podcast rows)

Every podcast row carries the episode context, regardless of family:

Field	Type	Required	Description
`podcast_name`	string	✓	Show name
`episode_title`	string	✓	Episode title
`episode_url`	string		Link to the episode — `null` when no canonical URL was resolvable
`episode_id`	string		Stable episode identifier. Join key across all rows extracted from the same episode; occasionally `null` on sponsorship rows
`published_at`	string (date)	✓	Episode publication date
`headline`	string	✓	One-line summary of the row, suitable for display

Appearance Fields

podcastAppearance / companyPodcastAppearance add:

Field	Type	Required	Description
`host`	string		Show host
`occasion`	string		Why the guest was on (product launch, book, event, series)
`co_guests`	array[string]		Other guests on the episode
`topics`	array[string]	✓	Topics discussed
`talking_points`	array[object]	✓	What the entity said, as `{point, quote}` pairs - `point` is the paraphrase, `quote` is verbatim from the transcript
`company_snapshot`	object	✓	The company as described in the episode: `what_they_do`, `org_facts[]`, `technologies_used[]`, `customers_mentioned[]`, `competitors_mentioned[]`
`transcript_source`	string	✓	Always `"transcript"` - extraction ran on the full episode transcript, not show notes
`recorded_at_estimate`	string (date or year-month)	✓	Estimated recording date, e.g. `"2026-07"` (episodes are often published weeks — or for reruns, years — after recording)
`date_confidence`	string	✓	`"exact"` or `"inferred"` — confidence in the recording-date estimate
`is_ad`	boolean	✓	`true` only for house ads (the show promoting its own product); regular rows are `false`

Claim Fields

The 65+ claim subtypes add:

Field	Type	Required	Description
`quotes`	array[string]	✓	Verbatim transcript quotes supporting the claim
`evidence`	string	✓	Convenience copy of the first quote
`speaker`	string	✓	Who said it
`speaker_relation`	string	✓	`own_company` (speaker talking about their own company) or `third_party` (someone else talking about the entity)
`timing`	string	✓	When the claim applies, as stated in the episode
`event_date`	string (date)		Present when a concrete date is stated for the event

Sponsorship Fields

podcastSponsorship adds:

Field	Type	Required	Description
`offer_texts`	array[string]	✓	Verbatim ad reads - includes promo codes, offer terms, landing pages
`is_ad`	boolean	✓	`true` only for house ads (the show promoting its own product); third-party sponsor reads are `false`

Example: Appearance Signal

{
  "signal_id": "dca86039-9c23-54d3-b9c2-00eacdb38baf",
  "signal_type": "podcast-contact",
  "signal_subtype": "podcastAppearance",
  "association": "contact",
  "detected_at": "2026-08-24T10:25:28Z",
  "contact": { "full_name": "Arundhati Bhattacharya", ... },  // standard contact object
  "company": { "name": "Salesforce", "domain": "salesforce.com", ... },  // standard company object

  "data": {
    "podcast_name": "CII Podcasts",
    "episode_title": "India@100: Powering the Future through Digital Acceleration ft Arundhati Bhattacharya",
    "episode_url": "https://shows.acast.com/ciipodcasts/episodes/india100-powering-the-future-through-digital-acceleration-ft",
    "episode_id": "55821260383",
    "published_at": "2026-06-05",
    "headline": "Arundhati Bhattacharya (President & CEO, Salesforce, South Asia @ Salesforce) on CII Podcasts",
    "host": "Rajan Navani",
    "topics": ["India's economic future", "Digital transformation", "AI adoption", "Cloud computing"],
    "talking_points": [
      {
        "point": "India must become a creator of technology, not just a consumer.",
        "quote": "I also want to see India not only as a consumer of technology, but actually as a creator of technology. India has long been a consumer and yet, you know, we are one of the most digitally empowered societies."
      }
    ],
    "company_snapshot": {
      "what_they_do": null,   // populated when the episode discusses the company itself;
      "org_facts": [],        // often sparse for interview-style appearances
      "technologies_used": []
    },
    "transcript_source": "transcript",
    "recorded_at_estimate": "2024-06",
    "date_confidence": "inferred",
    "is_ad": false
  }
}

Note: detected_at on podcast rows is when the weekly batch published. Use published_at and recorded_at_estimate in data for episode timing. Both podcast types are keyword-searchable via content_filters on the search endpoints (keywords clause only).

→ Full guide with subtype list, sponsorship examples, and transcript keyword search: Podcasts

Sample Files

Download example signals to explore the schema:

Browse Sample Files →

Note: Production data is delivered via GCS in timestamped folders. Each folder contains new signals since the last batch - not a full refresh. See Delivery for bucket access and refresh schedules.

Need custom samples or have questions? Contact [email protected].