“Job board” covers five very different things in this cluster. One is a documented API, one is a profile network that names the person who posted the job, one is the largest professional network reading as a signed-out visitor, one is a monthly comment thread, and one is a national board with its own salary notation. Treating them as interchangeable sources of rows is the first mistake, so this guide starts with how each is shaped.
Five sources, five shapes of data
| Source | How it is read | Sample measured | Primary price |
|---|---|---|---|
| Get on Board | Documented public API, no key | 934 jobs, whole board, 23 Sep 2026 | $0.45 / 1,000 jobs |
| Torre | Search plus job and profile pages | 330 open posts, 1 Oct 2026 | $3 / 1,000 posts |
| LinkedIn (18 LATAM countries) | Signed-out job search and job pages | 639 jobs, 8 countries, 1 Oct 2026 | $2 / 1,000 jobs |
| Hacker News Who is Hiring | Algolia HN Search API, no key | 848 company posts, Jul–Oct 2026 | $4 / 1,000 posts |
| elempleo.com (Colombia) | Site search and offer pages, no account | 1,000 offers, 30 Sep 2026 | $0.70 / 1,000 offers |
The samples are what make the numbers below comparable at all. A fill rate is a property of a source on a date, measured on a stated number of rows, and none of the five is a promise about the next run — why fill rates are published that way.
Salary is where they differ most
If the dataset is for salary research, the choice of board decides most of the answer before any analysis starts.
| Source | Salary present | Unit as delivered |
|---|---|---|
| elempleo.com | 75.9% of 1,000 | COP per month, parsed from a band |
| Get on Board | 41.2% of 934 | USD gross per month, converted by the board |
| Torre | 40.9% of 330 (minimum) | Currency and period as posted, plus hourly USD |
| Hacker News | 37% of 848 | Currency and period per band, as written |
| 2.7% of 639 | LinkedIn's published pay range |
The units matter as much as the rates. Get on Board converts every salary to a monthly USD figure rounded to 100 before it ever reaches the API, so a yearly offer looks monthly — the Get on Board API and its conversion. elempleo publishes bands in millions of pesos with a decimal comma, so “$2,5 a $3 millones” has to be parsed before it can be compared with anything — how elempleo is read. Hacker News posts list several bands and currencies in one post often enough that the Actor keeps every band, not just the first.
Which sources carry people
Two of the five sources describe companies only. Get on Board measured 0 of 934 jobs without a company, and elempleo only lets registered company accounts publish, 494 of 494 offers measured.
The other three carry named individuals. Torre shows the company person who posted a job on 97.1% of 138 company-published posts, and its recruiter mode is a list of people — what is collected about them. LinkedIn shows a job poster on 10.2% of 639 jobs. Hacker News delivers the username of each post's author, and its separate Who wants to be hired thread consists entirely of people looking for work — how that thread is handled.
Each Actor sets a default that withholds part of that. Torre excludes people located in the EU, EEA, UK or Switzerland; LinkedIn withholds posters on jobs located there; Hacker News hides personal-domain addresses and leaves the job-seeker thread off. The same reasoning drove the Wellfound Actor to skip company profile pages entirely — the data that Actor will not collect.
What each site says about machine access
| Source | robots.txt on 10 Oct 2026 | Terms or documentation |
|---|---|---|
| Get on Board | One comment line, no rules | Public API documented as open to anyone |
| Torre | Disallows /api/ and /search/jobs?*; search host has no robots.txt (404) | Lists scraping among prohibited uses |
| Disallow: / for every unnamed crawler | User Agreement prohibits scraping | |
| Hacker News | news.ycombinator.com sets a 30 s crawl delay; the Algolia API host has no robots.txt (404) | The Actor does not request news.ycombinator.com |
| elempleo.com | Paths read are allowed; comments prohibit automated extraction without written permission | Bars unauthorised commercial use |
Only Get on Board publishes an invitation. Two of the five — Torre's /api/ paths and LinkedIn as a whole — are read against an explicit disallow, which each Actor's page and guide states rather than leaves out — what LinkedIn's file says.
Combining them into one dataset
No job id survives across boards, so joining is by company. Torre, Hacker News and LinkedIn return a cleaned domain on 79.1%, 80.8% and 97.6% of their measured rows; Get on Board returns the company website on every job; elempleo returns an employer name and id, and nothing at all on the 31.5% of offers published as confidential. A domain join therefore covers four sources well and elempleo poorly, which is a property of elempleo rather than of the join.
Freshness differs too. Get on Board and elempleo are read live per run; Hacker News is a monthly thread that keeps growing for days after it opens; Torre's search mixes closed posts with open ones, 65 of 2,813 open in one measured query. For a freelance rather than salaried view of the same region, the Workana cluster covers it — Latin American freelance market data— and for startup roles with equity, the Wellfound guide.
What a month of data costs
At the primary prices, the whole of Get on Board is about half a dollar per pull, since the board held about 1,000 open jobs. One month of Hacker News company posts — 254 in September 2026 — is about $1.02. 1,000 LinkedIn jobs are $2.00 without company profiles. elempleo's whole national inventory, 20,876 offers on 30 September 2026, is about $14.60. Torre depends most on options: 100 posts with full pages are about $1.10, and recruiter profiles add $0.01 per person.
All five charge per delivered row and never for error rows or duplicates within a run, which is the property that makes those sums predictable — how pay-per-event pricing works.