Somoon Data documentation

Somoon provides hiring data for Korean listed companies: posting-level records for the full KOSPI and KOSDAQ universe, mapped to exchange identifiers, with point-in-time observation dates. Alongside postings, it offers company fundamentals from DART filings and monthly workforce flows from National Pension Service records.

There are three ways to work with the data: the Explorer and company profiles in the portal, the REST API, and bulk files of the full history.

Snapshot: 2026-10-21.

Quickstart

1. Sign in at data.somoon.ai with the credentials for your trial and accept the evaluation terms.
2. Create an API key under Account & API keys. Trial keys return sandbox data — synthetic companies and values with the production schema — so you can build and test your integration; subscription keys return live data.
3. Make your first request:

curl -s "https://data.somoon.ai/api/v1/metrics/postings?stock_code=005380&freq=month" \
  -H "Authorization: Bearer $SOMOON_API_KEY"

Or load the full history into pandas:

import pandas as pd
panel = pd.read_parquet("somoon_panel_weekly_2026-10-21.parquet")   # from Bulk downloads
universe = pd.read_parquet("somoon_universe_2026-10-21.parquet")
df = panel.merge(universe[["stock_code", "isin", "live_since"]], on="stock_code")
df = df[df.week_start >= df.live_since]          # live history only

Coverage

Postings held by year (posting date)

How far back the data goes, per company

Number of companies by the age of their earliest posting — counting all postings, and counting live postings only.

Datasets & fields

All datasets share stock_code and corp_code as join keys.

Job postings postings

Posting-level hiring data for KOSPI and KOSDAQ companies — one row per company and posting. Titles are in the original Korean. Posting body text and source URLs are not included.

Company universe & history universe

Every KOSPI and KOSDAQ company, including delisted ones, with ISIN, industry and per-company history depth. Join to the other datasets on stock_code.

Weekly hiring panel panel_weekly

Company × week counts of postings by posting date, split into live and backfilled — ready for backtests. Weeks start on Monday. Weeks with no postings have no row. Use postings_live for point-in-time work.

Company fundamentals (DART) fundamentals

Headcount, average salary by gender, average tenure and key financials from annual DART filings — latest available year per company. Monetary values in KRW. Data year is given per company because filing cycles differ.

Monthly workforce flows (NPS) workforce

Monthly hires, leavers and insured headcount per company, derived from National Pension Service workplace records. Based on the company's registered NPS workplace(s); companies with workplaces registered under other entities may be partially covered.

Methodology

Universe

All companies listed on KOSPI and KOSDAQ, synchronized weekly with DART. Companies that delist are kept with their history.

Collection

Publicly available job postings are collected every week for the full universe from company career sites and major Korean job platforms, then normalized and mapped to the listed company.

History before collection

Many platforms keep closed postings online. When a company is first collected, postings still available are included and dated by the date published in the posting. Such postings have a first_seen_date later than their posted_start_date.

De-duplication

A posting re-observed on later collections is kept once per company. Group-wide postings that list several affiliates appear under each affiliate with the same posting_id; companies_sharing_posting gives the count. Count distinct posting_id for unique postings.

Identifiers

KRX stock code and DART corporate code come from DART. ISIN is derived for the common share from the stock code (KR7 + code + 00 + ISO 6166 check digit); preferred shares and other share classes have different ISINs.

Dates and collection type

posting_date is the published start date where the posting states one, otherwise the date we first observed it. Every posting carries a collection_type: live when we observed it within 30 days of publication, backfilled when it was published earlier and gathered when we began collecting the company. first_seen_date records when we actually observed each posting and is never revised.

Fundamentals and workforce

Fundamentals are the latest annual figures filed with DART; the data year is given per company. Workforce flows come from National Pension Service workplace records and depend on how a company registers its workplaces.

Point-in-time & backtesting

Backfilled postings extend history, but they were gathered after the fact. For a signal that only uses information available at the time, use live postings.

For live postings that carry a published start date (28,247 postings), the gap between publication and our first observation was: median 4 days, 90th percentile 18 days; 70.2% were observed within 7 days.

Trial & subscription

The trial is designed for evaluation: every aggregate and the full weekly panel, with posting-level detail available in samples.

Limits are applied by the service to the portal and the API alike. To extend your trial or discuss a subscription, use the chat button or email bkpark@seeso.kr.

API reference

Base URL https://data.somoon.ai. Authenticate with Authorization: Bearer <API key>; keys are created in the portal. Responses are JSON. List endpoints return total and a data array and accept limit / offset. Errors return a non-2xx status with {"error": ..., "message": ...}; plan limits return 403 with "error": "plan_limit". Each response carries X-Somoon-Environment: live or sandbox.

GET /api/v1/coverage — Coverage summary

Universe size, posting totals and history depth by year for the current snapshot. Try it →

GET /api/v1/companies — Search companies

Find companies in the KOSPI/KOSDAQ universe by stock code or name (English or Korean). Try it →

GET /api/v1/companies/{stock_code} — Company profile

Identifiers, industry, hiring summary and latest DART fundamentals for one company. Try it →

GET /api/v1/postings — Query postings

Posting-level records with filters. Subscriptions can export the full filtered result via /api/v1/postings.csv. Try it →

GET /api/v1/metrics/postings — Posting counts over time

Unique postings (split live / backfilled) and number of hiring companies per week, month or year — same filters as /postings. Try it →

GET /api/v1/workforce — Monthly workforce flows

Monthly new hires, leavers and insured headcount from National Pension Service records. Try it →

GET /api/v1/fundamentals — Company fundamentals

Headcount, average salary by gender, tenure and financials from annual DART filings. Try it →

Bulk delivery

During the trial, every dataset is available as a full-history file in CSV (UTF-8 with BOM) and Parquet (zstd) under Bulk downloads. After the trial, weekly full refreshes are delivered by file download or to your S3 bucket / SFTP.

Data FAQ

Answers to the questions in a standard data-sourcing questionnaire.

Compliance & sourcing

• No personal data. No information about applicants or individual employees is collected or provided.
• Public information. Job postings are publicly accessible online. Fundamentals come from public DART filings; workforce flows from public National Pension Service data.
• Trial scope. Posting body text and source URLs are excluded from the trial.
• Access records. Sign-ins, acceptance of the evaluation terms, downloads and API calls are recorded.

Questions from compliance or legal teams are welcome — contact bkpark@seeso.kr.

Changelog