Framework — Creative Labor Observatory

ACTIVE OPENINGS ↗

—

Observed, not total U.S. vacancies

STUDIOS HIRING ⌘

—

Distinct recruiting-board identities

JUNIOR / INTERN ↗

—

Open postings, not hires

LAYOFF VOLATILITY ◷

N/A

Needs 8+ weeks of WARN history

00A / INDUSTRY WEEKLY

Tracked-board hiring history

—

Loading upstream weekly indices…

Gaming Hiring Index and concentration from the Job Tracker (industry-wide tracked boards). Not identical to our U.S. snapshot totals.

00B / US OBSERVED DAILY

Daily hiring observations

Collecting

History is collecting. Trends appear after 14 observation days.

UTC calendar-day snapshots of tracked U.S. listings. Observation date is collection time, not posting time. A disappeared listing is not a layoff.

01 / DISTRIBUTION

Hiring by discipline

Current snapshot

Awaiting first source import

Counts of active advertisements by classified job family; not headcount.

02 / SENIORITY

Who's getting hired?

Awaiting first source import

Seniority is inferred or reported by the upstream source.

03 / MARKET CONCENTRATION

Most active hiring studios

Awaiting first source import

04 / RESEARCH STATUS

What can we know?

Postings are observable signals of hiring intent. They are not verified hires, and disappeared listings are not layoffs.

CONNECTED*

Active games listings*After successful ingestion

CONNECTED*

WARN cohort CA/NY/TX/WA*Coverage incomplete by design

CONNECTED*

BLS QCEW & Census BDS*Separate NAICS baselines

PLANNED

NCES graduate completionsField-of-study classification needed

05 / OFFICIAL BASELINES

Government industry context

Not studio counts

Baseline series not published yet.

BLS QCEW employment and Census BDS establishments use government NAICS classifications. They are never added into observed job or studio totals.

LISTINGS / CURRENT OBSERVATIONS

Open positions

—

Awaiting first source import

GEOGRAPHY / OBSERVED CITIES

Where roles are named

—

Pins mark the U.S. city named on a listing (Census place centroid), not a verified office address. Nearby pins merge within 50 screen pixels when zoomed out. Listings without a resolvable city stay off the map.

STUDIO / OBSERVATIONAL PANEL

Tracked studio hiring

Derived from job listings

This lists recruiting-board identities with active positions, not a census of all studios or business establishments.

Awaiting first source import

WARN / STATE NOTICES

Layoff filings in cohort states

CA · NY · TX · WA

WARN notices are legally required filings above size thresholds. This cohort covers California, New York, Texas and Washington only. Missing states are not zero. Notice-date and effective-date series are kept separate. Studio name matches are observational.

ROLLING MEAN

N/A

12-week window

ROLLING SD

N/A

Sample standard deviation

ROLLING MAD

N/A

Median absolute deviation

TRACKED MATCHES

—

Name overlap with job boards

Volatility stats stay blank until at least 8 non-null weeks exist in the rolling window.

WARN weekly chart loading…

RECENT NOTICES

Latest filings

No WARN notices published yet.

CONTEXT ONLY

QCEW beside WARN

Baselines not loaded.

WARN filings and QCEW employment are not commensurate. Shown together only as industrial context.

MODEL / DESCRIPTIVE BASELINES

Trajectory models

—

Descriptive baselines only. Not a forecast of hires, layoffs, or employment.

GHI ROLLING MEAN

N/A

8-week window

GHI ROLLING SD

N/A

Sample SD

HOLDOUT MAE

N/A

One-step naive, last 4 weeks

HOLDOUT MAPE

N/A

Percent error when defined

Trajectory series loading…

Solid = observed Hiring Index. Amber dashed = naive carry-forward. Teal dashed = mean week-over-week drift. Holdout evaluates one-step naive on the last 4 observed weeks.

SECONDARY

WARN weekly trajectory

WARN trajectory loading…

SOURCE PROVENANCE

Methods before metrics.

Framework deliberately distinguishes verified source observations, official published statistics and model-derived estimates. A blank chart is preferable to an invented number.

Gaming Industry Job TrackerLive feed · Weekly indices history · Daily US history in Parquet/R2 · Attribution required

Documentation ↗

State WARN open data (CA, NY, TX, WA)Connected cohort · Notice date vs effective date separated · Coverage metadata published

Coverage JSON ↗

BLS Quarterly Census of Employment and WagesConnected baseline · NAICS 5112 / 5121 / 5415 · Never merged into observed totals

Documentation ↗

U.S. Census Business Dynamics StatisticsConnected baseline · Establishments and job flows by industry

Documentation ↗

Trajectory modelsDescriptive · Rolling mean/SD/MAD · Naive + drift paths · Holdout MAE/MAPE · Not hire forecasts

trajectories.json ↗

NCES IPEDSPlanned · Annual educational awards, not direct placement rates

Documentation ↗

Data contract

GET /data/jobs.json
GET /data/series/jobs_daily.json
GET /data/series/hiring_index.json
GET /data/series/concentration.json
GET /data/series/trajectories.json
GET /data/series/warn_weekly.json
GET /data/series/bls_qcew.json
families: observed.* | upstream.index | baseline.* | model.trajectory
trend_ready / stats.ready gate blank UI

The jobs cache is updated atomically. Failed imports never overwrite a valid prior snapshot. Daily US history is restored from R2 before each append in CI. Weekly industry indices come from the upstream Job Tracker API and are labeled separately from our US snapshots. Disappeared listings are not layoffs. Trajectories are descriptive baselines, not predictions of employment outcomes.

Creative labor, measured.

Tracked-board hiring history

Daily hiring observations

Hiring by discipline

Who's getting hired?

Most active hiring studios

What can we know?

Government industry context

Open positions

Where roles are named

Tracked studio hiring

Layoff filings in cohort states

Latest filings

QCEW beside WARN

Trajectory models

WARN weekly trajectory

Methods before metrics.

Data contract