Tony Wang10 min readMore Than 16,000 Jobs Say Only ‘2 Locations’
‘2 Locations’ leads the Jobs location facet at 16,306. Multi-site placeholders total 29,411—about 5.6% of open postings. Not a city ranking.
Ask an ATS board where a job is and you often get a number, not a place. In Crawlora’s Jobs by_location facet, the #1 location string is “2 Locations” — 16,306 open postings — larger than the country key “United States” (3,599) and larger than every named city in the same top-100 window. “3 Locations” through “8 Locations” keep stacking. Add them up and 29,411 open roles refuse a single city before you start merging spellings of London or Bengaluru.
This post is about that structure: multi-site placeholders, free-text remote, and what country totals look like only after you throw the unresolvable rows out. It is not a second “where the jobs are” city list — that story is already Bengaluru vs San Francisco.
“2 Locations” is bigger than “United States”
Sort the top location strings by count and the first row is not a labor market you can put on a map.
“2 Locations” (16,306) is about 4.5× the country key “United States” (3,599) in the same facet window. That does not mean the US has few jobs — most US roles sit under city strings (New York, NY, San Francisco, CA, …) or never appear in the top 100. It means the modal location label among high-volume raw strings is a multi-site opaque count, not a city name.
Enterprise ATS UIs (Workday is the usual source of N Locations language) collapse multi-site reqs into a single posting with a placeholder location. For hiring managers that is convenient. For anyone ranking cities or countries from the raw field, it is a trap.
Multi-location placeholders: the full stack
Exact N Locations keys in the top 100:
| Location string | Open postings | % of top-100 mass | % of open (provider sum) |
|---|---|---|---|
| 2 Locations | 16,306 | 15.6% | 3.1% |
| 3 Locations | 5,390 | 5.1% | 1.0% |
| 4 Locations | 3,328 | 3.2% | 0.6% |
| 5 Locations | 1,923 | 1.8% | 0.4% |
| 6 Locations | 1,235 | 1.2% | 0.2% |
| 7 Locations | 731 | 0.7% | 0.1% |
| 8 Locations | 498 | 0.5% | 0.1% |
| All 2–8 Locations | 29,411 | 28.1% | 5.6% |
29,411 open postings — 5.6% of the crawl by provider sum — have no city and no country in the location string. Inside the top-100 location mass, multi-loc is 28% of everything you would naïvely rank. That is why the Bengaluru post excluded these rows from city tables: folding “2 Locations” into any hub would invent geography.
If your pipeline ranks cities from raw ATS JSON without a multi-loc filter, your #1 “city” may be a digit.
Free-text remote is not a country either
Before country rollups, peel off workplace language that looks like a place:
| Pattern (top-100) | Open postings |
|---|---|
Remote | 3,920 |
Remote job | 1,218 |
US-tagged remote forms (Remote (United States), Remote - US, US Remote, …) | 4,415 |
| Free-text remote subtotal | ~9,553 |
These rows are not the remote=true boolean (a companion post covers the remote flag). They are parallel encoding: location field used as workplace type. Country totals below exclude them so “United States” is not double-counted via Remote (United States) as if it were a geo rollup of on-site US hiring.
Country rollup — only after exclusions
Hand-assign the remaining top-100 keys to countries (city variants merged into ISO-style buckets). Scope: resolvable keys in the top-100 window after dropping multi-loc, free-text remote, and non-geo noise (~1,946 hospital/network labels). This is not “the US has 6.5% of all jobs.”
| # | Country / territory | Open postings (scoped) | Includes (examples) |
|---|---|---|---|
| 1 | United States | 34,000 | City forms + coarse ‘United States’ |
| 2 | India | 8,577 | Bengaluru/Bangalore, Hyderabad, Pune, Mumbai, ‘India’ |
| 3 | United Kingdom | 6,900 | London variants + ‘United Kingdom’ |
| 4 | Germany | 2,979 | Berlin, Hamburg, Munich/München |
| 5 | Singapore | 2,552 | Singapore city-state keys |
| 6 | Canada | 1,989 | Toronto variants + ‘Canada’ |
| 7 | Mexico | 1,090 | Mexico + Guadalajara |
| 8 | Philippines | 911 | Coarse country key |
| 9 | Greece | 669 | Athens / Attica |
| 10 | Japan | 621 | Tokyo |
| 11 | Saudi Arabia | 617 | Riyadh |
| 12 | Spain | 540 | Madrid |
| 13 | Hong Kong | 523 | Hong Kong |
| 14 | France | 517 | Paris |
| 15 | UAE | 479 | Dubai |
| 16 | South Africa | 469 | Coarse country key |
| 17 | South Korea | 416 | Seoul |
United States leads the scoped rollup because US city strings dominate the high-volume tail of the top 100 once multi-loc is removed. India is second even before you leave the top-100 window — consistent in direction with the Bengaluru hub story, but measured here as country-bucket mass among large location labels, not as a city ranking.
Singapore appears as both a city-scale hub and a country-scale key; we keep it as one country/territory bucket so it is not double-counted as “city and country.”
How this differs from the Bengaluru post
| Question | Bengaluru post | This post (G1) |
|---|---|---|
| Unit | City hubs (merged variants) | Location-field structure + country rollup |
| Multi-loc | Excluded from ranks (~76k in older full-corpus note) | Headline finding (29,411 exact keys in this pull) |
| Ranking claim | Bengaluru > SF+Boston | “2 Locations” > country key “United States” |
| Risk if mixed | Inflated cities | Fake countries from remote strings / multi-loc |
Same dataset, different intent. If you need “which city,” use the city post. If you need “how broken is location,” use this one.
What this is not
- Not a full country census of ~525k open postings. Only top-100 location keys are assigned; most mass sits in the long tail.
- Not a city ranking. London and New York appear only as inputs to country buckets.
- Not remote market share. Free-text remote rows are excluded from countries; boolean
remote_openis a separate honesty story. - Not proof enterprise hiring is “multi-site only.” It is proof multi-site labels are first-class volume in the raw field.
Who this helps
Analysts / data eng: filter ^\d+ Locations?$ (and sibling multi-site phrases) before any city or country aggregate. Treat location as a dirty enum.
Product builders: if you expose “jobs by country,” document multi-loc and remote-string exclusions or you will ship a dashboard that ranks “2 Locations” as a place.
Researchers citing the series: multi-loc share and city ranks must never share a denominator without the exclusion callout.
Everything above is re-queryable from the same facets endpoint. When a re-crawl moves “2 Locations,” update the floor — do not invent cities for it.
Inspect live location strings across 14 ATS providers
Search open roles, facet by location, and see multi-site placeholders next to real cities — same REST surface as this study. 2,000 free credits a month, no card.
Frequently asked questions
What is the most common location string on public ATS job postings?
In Crawlora’s 2026-07-19 top-100 by_location facet, the largest raw string is ‘2 Locations’ at 16,306 open postings — ahead of ‘United States’ (3,599) and every city key in that window. It is a multi-site placeholder, not a city.
How many jobs have no single city in the location field?
Exact ‘N Locations’ placeholders from 2 through 8 Locations sum to 29,411 open postings — about 5.6% of roughly 525,000 open postings by provider sum, and about 28% of the mass inside the top-100 location keys.
Which countries lead after multi-location rows are excluded?
Among resolvable top-100 location keys after excluding multi-loc placeholders, free-text remote strings, and non-geo noise, a hand rollup is led by the United States (~34,000), India (~8,600), the United Kingdom (~6,900), Germany (~3,000), and Singapore (~2,600). That is scoped to the top-100 window, not a full-corpus country census.
Is ‘2 Locations’ larger than US hiring?
No. ‘2 Locations’ is larger than the coarse country key ‘United States’ in the same facet window, but most US roles sit under city strings (New York, San Francisco, …) or outside the top 100. The finding is about label structure, not total US volume.
How is this different from the Bengaluru city ranking post?
That post ranks cities after merging variants and excluding multi-loc rows. This post treats multi-location placeholders as the headline finding and builds a country rollup only after those exclusions. Same Jobs dataset, different unit: city hubs versus location-field structure.
Does free-text ‘Remote’ count as a country?
No. Generic and US-tagged remote location strings (~9,550 in the top-100 window) are workplace language and are excluded from country totals. The remote=true boolean is covered separately in the remote-flag honesty post.