Home / Method

Method

What is counted, where the rows come from, what was dropped and why, and what the counts are checked against.

Counts of 2026-09-15

1. What is counted

A row is editorial when its job title matches one of these phrases: bureau chief, news director, managing editor, editor-in-chief, sub-editor, correspondent, columnist, anchor, presenter, news host, newsreader, photojournalist, journalist, reporter, critic, reviewer, news producer, newsroom producer, editorial producer, producer-reporter, staff writer, feature writer, editor, editorial, and when its title matches nothing on the stoplist.

The stoplist is applied first, so "marketing editor" and "product editor" are out. It covers sales, marketing, brand, growth, advertising, ad sales and ops, media buying, account management, HR, recruiting, talent, payroll, people ops, engineering, developers, devops, QA, software, data engineering and science, IT support, product roles, content strategy, content marketing, SEO, copywriting, content and technical writing, grant, proposal, ghost and bid writing, PR and communications, finance, accounting, legal counsel, paralegals, facilities, security guards, drivers, customer success, support and service, subscriptions, circulation, teachers, professors, lecturers and students.

Three exclusions are worth knowing about. A bare producer is not editorial: media companies are full of TV and radio production staff, so only news, newsroom and editorial producers and producer-reporters count. Writers count only as staff writer and feature writer, which drops 10,767 rows whose title is bare writer or senior, contributing, lead, freelance, news, sports, science, business or political writer. Some of those are real newsroom jobs; the number is recorded so the rule can be loosened and rerun. Video, photo, copy and audio editors are kept and tagged.

Roles are reporter, editor, correspondent, columnist, anchor or presenter, producer, critic and other editorial. The first match in the include list wins, so bureau chief beats reporter and managing editor beats editor.

Beats

A lexicon of 26 beats is run over the job title and the profile headline together. A row can carry several beats and is counted under each of them. 31,773 of the 127,602 journalists on file, 24.9%, carry any beat at all, which is why every beat figure on this site says titled.

The first version of the lexicon put education second overall. It was matching university in headlines like "Student at University of Washington", not education reporters. Entertainment was matching tv and television in the names of local stations, and technology was matching digital in "digital media editor". Those tokens were removed. Fixing it cut the share of rows with a beat by about a third.

Seniority

From the words in the title, first match wins: chief, executive, deputy, senior, staff, else unknown. Most titles carry no seniority word, so most rows read unknown.

2. Where the rows come from

Three tables: an outlet universe, the people at those outlets, and Wikidata's journalists.

• The outlet universe is a vendor company dataset cut to media industries, a hand-listed set of 125 outlets, the Wikidata news-media-organisation subtree, and every employer of a Wikidata journalist that has a website.
• The people rows come from a vendor people dataset, joined to the outlet universe on the company domain. Titles and headlines are text the person wrote on their own professional profile.
• Wikidata contributes 208,997 journalists, of whom 21,064 are living and have an employer recorded.

Every request went through a rotating proxy with browser headers. No request was made from the machine's own IP, and no outlet site was crawled for this pull.

The outlet universe

A domain in several sources is attributed to the first of the hand list, the Wikidata subtree, the vendor cut, then Wikidata employers.

Of the hand-listed outlets, 59 are inside the vendor's core media cut, 64 are filed under another industry and 2 have no company record at all: 66 of 125, 52.8%, are missed by the industry cut alone. TechCrunch is filed under aviation and aerospace, WIRED under graphic design, The Verge under publishing, Politico under online audio and video media, the Wall Street Journal under financial services and USA Today under restaurants.

Outlet type

Every Wikidata instance-of label on an outlet's domains is mapped into one of five buckets, newspapers, broadcasters, online outlets, wire agencies and magazines, and the buckets are counted: the bucket with the most labels wins. A domain with no Wikidata entity falls back to its vendor industry label. Where two buckets tie and the industry label does not break the tie, the more specific bucket wins, in the order wire agency, newspaper, magazine, broadcaster, online. An outlet the rule cannot type is left off every outlet table.

Outlet country

The country of an outlet is the one its own rows in the outlet universe agree on, with the hand-listed and Wikidata rows winning over the vendor rows, which carry one country per company record and disagree across a group's regional entries. Where the universe carries no country at all, the country most of that outlet's own journalists sit in stands in. A vendor country string is only read for the ISO country code it names. The name printed comes from a fixed English table, so no vendor spelling reaches a page.

3. The funnel

Every number on this site is drawn from the published base.

The first drop is the platform domains. The vendor stores a link-in-bio, social or freelance-marketplace URL as the company website for thousands of small media companies, and because the join key is the domain, every one of those people lands in a single bucket: 626,791 rows on one link-in-bio service, 499,397 on one link shortener, 207,481 on one professional network. Those domains are listed in the classifier and their rows are kept in the file, so the decision can be reversed.

The second drop is the outlets that are not news outlets. Wikidata's employer source is every employer of a Wikidata journalist, and journalists work at universities, academic publishers and government agencies, so those land in the universe too and their staff carry titles the classifier reads as editorial. The two largest single domains in the unfiltered editorial table are an academic publisher with 2,910 rows and another with 2,115. A domain is kept when its source is the vendor media cut or the hand list, or when its Wikidata instance-of label is on an explicit news-outlet list: newspaper, news agency, news website, news media, news channel, programme, broadcast or portal, television channel, station, network or broadcaster, radio station, network or broadcaster, magazine, tabloid, wire service, press agency, media company, mass media. Publishing house, book publisher, academic journal, university press, newsletter, website, organization, public company and university are not on it.

Profiles were fetched between 2026-07-22 and 2026-09-15, median 2026-08-05.

4. Sanity checks

Three checks. Both figures are given each time.

The BBC

This file has 3,935 journalists at the BBC, counting bbc.co.uk, bbc.com and bbcstudios.com as one outlet. The published figure is 5,500 journalists, from the BBC News infobox on Wikipedia, cited there to a BBC Studios page dated 10 December 2023. The file is at 71.5% of the published figure.

The vendor claims

This file has 160,593 editorial rows off the platform domains, 127,602 of them at outlets typed as news media. Press Hunt's own homepage title reads that it is a journalist database of 500k+ journalists. The file is at 32.1% of that claim. The two count different things: these rows are current employment records at companies filed as news media, and a PR database counts profiles including freelancers, bloggers and podcasters with no employer. Muck Rack's own site is unreachable on every route we have, including its robots.txt, so its database size is unverified.

A masthead

TechCrunch names 43 people on its staff page, 26 of them under Editorial and Editorial Contributors. This file has 24 editorial rows at techcrunch.com, and 13 of the 26 masthead names are among them. Several of the rest are former staff.

5. What is published about a person

A listed row carries the full name, the job title as the title field reads, the outlet, the country, the role, the seniority and the beat. The profile headline is user-written text and is never published. There is no contact detail of any kind, no profile URL, no page per person and no person-level download. A row with a blank name, title, outlet or country is dropped from every list.

A segment share or a list page is published only when its base reaches 40 rows. Below that the page prints n/a. A country page is published at 200 rows or more.

6. Limitations

• The vendor's industry labels are wrong often enough to matter: 66 of the 125 hand-listed outlets, 52.8%, sit outside its core media cut.
• Wikidata's employer source is every employer of a Wikidata journalist, which includes universities, academic publishers and government agencies. The news-media filter keeps those out of the published base.
• Outlet aliases cover 395 domains, not all 177,692. Outside the largest outlets, a per-outlet count is a per-domain count. Parent-company domains are deliberately not merged into their titles, so a title's own domain undercounts its newsroom, and the parent companies are left off the outlet tables entirely.
• The country on a row is the country on the person's own profile. It disagrees with the outlet's country often: a BBC row can carry US.
• Titles and headlines are written by the person, not by an employer. People write aspirational titles, keep old employers and add keywords to headlines. The beat lexicon reads those headlines, so the same text decides the beat.
• A fetch date records when the profile was scraped. It does not record when the person last held the job, so departed staff are still present.
• Coverage is thin outside the large outlets. Of the 177,692 domains in the universe, 8,800 have an editorial row at all, 3,597 have five or more, 1,392 have twenty or more and 262 have a hundred or more. Most local outlets in the universe have no editorial row at all.
• Wikidata covers journalists across all of history and skews to the dead and the notable. About 15% of them have an employer recorded.
• No beat data exists in either vendor source. Beats are inferred from title and headline text only.
• Unverified: Muck Rack's actual database size, any first-party BBC headcount page, and whether the vendor's size band for an outlet refers to the outlet or to its parent group.

7. Rebuilding the counts

The classifier, the alias table and the export script are in the repository. python3 tools/export.py rebuilds every file this site reads from the raw tables.