Intent Audience Fields

Complete reference for every field returned in Intent Audience parquet exports. Each audience export contains up to 125 fields across intent scoring, contact, company, and demographic categories.

How Intent Audiences Work

1. Create an audience via the Intent Audiences API with topic filters
2. Poll until the audience task completes
3. Download the resulting parquet files
4. Each row represents one person matched to one intent topic

A person appears in multiple rows if they match multiple topics in the audience.

Intent Fields

These fields define the intent match and are always present (100% populated) on every row.

Score Details

• score: Categorical label -- high or medium
• perc_score: Numeric percentile rank (higher = stronger intent signal). Range: 51-98.
• ts: Date-level timestamp. The time portion is always midnight.

Identity Fields

Email and identity resolution data. Present on every row.

Email Hashes

Multiple hash algorithms and case variants for matching against your own data:

Additional Elixir-sourced hashes are available with the emails_ prefix: emails_md5_lc_hem, emails_sha1_lc_hem, emails_sha256_lc_hem, and their _uc_ variants.

Person Fields

Contact-level data from the identity graph.

Name

Email

Email Validation

Phone Numbers

All phone fields use E.164 format (e.g., +15551234567). Multiple numbers are comma-separated. Each phone field has a matching DNC (Do Not Call) field with positionally-aligned y/n values.

DNC alignment example:

phones: "+15551234567, +15559876543"
phones_dnc: "n, y"

The first number is not on the DNC list; the second is.

Professional Profile

education_history Example

{
  "name": "Stanford University",
  "url": "https://www.linkedin.com/school/stanford-university",
  "extraction_order": 0,
  "extraction_date": "2025-05-04T16:41:23Z"
}

Social Profiles

Personal Address

Demographics

Company Fields

Firmographic data from the identity graph.

Reading Parquet Files

See our Reading Parquet Files guide for language-specific examples.

Quick Start with DuckDB

-- Read all parts of an audience export
SELECT * FROM read_parquet('audience_part-*.parquet');

-- Filter to high-intent contacts
SELECT first_name, last_name, email, company_name, job_title, score, perc_score
FROM read_parquet('audience_part-*.parquet')
WHERE score = 'high'
ORDER BY perc_score DESC;

-- Unique contacts across all topics
SELECT DISTINCT ON (email) email, first_name, last_name, company_name
FROM read_parquet('audience_part-*.parquet');

Quick Start with Python

import pandas as pd

df = pd.read_parquet('audience_part-0000.parquet')

# High-intent contacts with company data
leads = df[(df['score'] == 'high') & (df['company_name'].notna())]
leads[['first_name', 'last_name', 'email', 'company_name', 'job_title', 'perc_score']]

All Fields (125 total)

address_id, age_range, business_emails, category, company_address, company_address2,
company_city, company_country, company_description, company_domain, company_employee_count,
company_employee_count_range, company_id, company_id_right, company_industry,
company_linkedin_url, company_naics, company_name, company_name_history, company_phones,
company_phones_dnc, company_related_domains, company_revenue_range, company_sic,
company_state, company_total_revenue, company_zip_code, current_business_email,
current_business_email_validation_date, current_business_email_validation_status,
department, department_2, department_raw, direct_numbers, direct_numbers_dnc, domain_lc,
dpv_code, education_history, email, email_last_seen, email_validation_status, emails,
emails_md5_lc_hem, emails_md5_uc_hem, emails_sha1_lc_hem, emails_sha1_uc_hem,
emails_sha256_lc_hem, emails_sha256_uc_hem, facebook_url, first_name, first_uuid_norm,
flattened_uuids, gender, github_url, has_children, headline, hem, income_range_lc,
inferred_gender, inferred_gender_unisex, inferred_years_experience, invalid_emails,
is_email_business, is_email_personal, is_homeowner, is_international, is_married,
is_profile_b2b, is_profile_b2c, job_functions, job_title, job_title_history,
job_title_normalized, last_name, last_updated, linkedin_url, md5_lc_hem, md5_uc_hem,
middle_name, mobile_phones, mobile_phones_dnc, mobile_phones_validation_date,
mobile_phones_validation_status, net_worth, perc_score, personal_address,
personal_address_2, personal_city, personal_country, personal_country_alpha2,
personal_country_alpha3, personal_country_numeric, personal_email,
personal_email_validation_status, personal_emails, personal_phones, personal_phones_dnc,
personal_state, personal_state_code, personal_timezone, personal_zip, personal_zip4,
phones, phones_dnc, photo_url, primary_contact_emails, profile_pid_all, score,
segment_type, seniority_level, seniority_level_2, seniority_level_raw, sha1_lc_hem,
sha1_uc_hem, sha256_lc_hem, sha256_uc_hem, social_connections, subcategory,
subdepartments, topic_id, topic_id_prefix, topic_name, ts, twitter_url, valid_emails

Data Formatting Notes

• Text fields are lowercase. Names, addresses, industries, and most string values are stored in lowercase. Apply your own casing for display.
• Boolean-like values are strings. Fields like has_children, is_homeowner, and is_married return "y" / "n" (not true / false). is_email_business and is_email_personal return "true" / "false".
• JSON fields are stringified. education_history is JSON encoded as a string. Parse it in your application.
• Null fields appear as null. If a field has no value for a given row, it will be null in the parquet file. Handle missing values in your code.
• Dates use ISO 8601. ts is a timestamp. Date-only fields use YYYY-MM-DD.

Related

• Intent Audiences API -- Create, poll, sample, and download intent audiences
• On-Domain Event Fields -- Field reference for pixel event data
• Reading Parquet Files -- Language-specific parquet reading examples
• Building Audience Filters -- Use fields in audience filter rules
• Person-Level Intent -- Understanding person-level intent data