Computrabajo salaries in Python: half the postings show pay, and A convenir comes out as 0
We read 36 Computrabajo search pages in six countries and 24 job pages. 49 % of 706 postings showed a salary, always as one monthly figure, from 21 % in Argentina to 68 % in Colombia. Where the pay is A convenir, the structured data says 0. How to read the pages, and what to clean before you take a median.
Computrabajo is a job board with one site per Latin American country: mx.computrabajo.com for Mexico, co. for Colombia, then pe., ar., cl., ec. and more. Its search pages are plain HTML. Each posting is an <article> with the title, the company, the place, the salary when the employer shows it, a remote or hybrid tag and the age of the ad, so no browser and no login are needed. If you were looking for a Computrabajo API: robots.txt closes /Ajax/ and /_services/, so we read the pages.
All six sites serve the same robots.txt. It leaves the search pages (with ?p= for later pages) and the posting pages open, and closes /ofertas-de-trabajo/ URLs with filter parameters such as sal= and pubdate=. It asks for no pause between requests; we waited 3.5 seconds.
On 3 November 2026 we read the first two pages of three searches, contador (accountant), vendedor (salesperson) and programador, in Mexico, Colombia, Peru, Argentina, Chile and Ecuador. That is 36 pages and 706 postings. Then we opened 24 of the postings, two with a salary and two without in each country.
The short version#
- About half show a salary: 344 of 706 postings, 49 %. It went from 21 % in Argentina to 68 % in Colombia, and programador ads showed it least, 85 of 229.
- Always one monthly figure. All 344 salaries were a single number marked (Mensual), none a range. How the number is written changes by country, and
$stands for four currencies. - No salary is not empty in the data. The posting page says A convenir (to be agreed), and its schema.org
baseSalarysays 0 a month: 11 of 11 pages. Read it as it is and a hidden salary becomes zero. - Clean before you take a median. In Argentina 9 of the 25 shown salaries were same-digit figures such as 11,111. In Chile 28 of 65 were exactly 1,000,000 pesos, from two companies.
- "Hace 5 horas" is the last update, not the posting date. 77 % of the rows said minutes or hours ago; on the job pages,
datePostedwas often days older.
What the salaries say#
The median of the monthly salaries shown in the contador search, first two pages, in local currency:
- Mexico: MXN 17,750, from the 32 of 40 ads that showed pay.
- Colombia: COP 2,556,000, from 31 of 40.
- Peru: PEN 1,630, from 11 of 40.
- Argentina: too few to say, 4 of 37.
- Chile: CLP 800,000, from 24 of 40.
- Ecuador: USD 575, from 22 of 40.
These are medians of what the ads show, from the ads that show it. Peru's rests on 11 ads. Argentina had 4, one of them a same-digit figure, so we give no median.
The share of postings with a salary and with a remote or hybrid tag, all three searches together (120 postings a country; 117 in Argentina after repeats, 109 in Ecuador, where programador found only 29):
- Mexico: salary 65 %, remote or hybrid 21 %.
- Colombia: salary 68 %, remote or hybrid 16 %.
- Peru: salary 29 %, remote or hybrid 12 %.
- Argentina: salary 21 %, remote or hybrid 27 %.
- Chile: salary 54 %, remote or hybrid 45 %.
- Ecuador: salary 55 %, remote or hybrid 10 %.
Most of those tags say hybrid (Presencial y remoto): 132 hybrid against 23 remote. Chile's 45 % comes largely from two companies, which gave 33 of its 54 remote or hybrid postings. The rest carry no tag, so count them as not stated, not as on-site.
The heading of each search page gives the count. Vendedor found 36,721 postings in Mexico and 433 in Ecuador; programador found 1,029 in Mexico and 29 in Ecuador.
The code#
import html, json, re, urllib.request
UA = {"User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 "
"(KHTML, like Gecko) Chrome/130.0.0.0 Safari/537.36",
"Accept-Language": "es-419,es;q=0.9"}
# Four of these countries write "$", Peru "S/." and Ecuador "US$" after the number: take the currency from the country.
CURRENCY = {"mx": "MXN", "co": "COP", "pe": "PEN", "ar": "ARS", "cl": "CLP", "ec": "USD"}
MODE = {"Remoto": "remote", "Presencial y remoto": "hybrid"} # no tag: the ad does not say
def get(url):
req = urllib.request.Request(url, headers=UA)
return urllib.request.urlopen(req, timeout=30).read().decode("utf-8", "replace")
def text(fragment):
return html.unescape(re.sub(r"<[^>]+>", " ", fragment)).split()
def money(salary):
"""'$ 1.750.905,00 (Mensual) + Comisiones' -> (1750905.0, 'Mensual', True). Mexico writes 12,000.00, the
others 12.000,00: the last separator before two digits is the decimal one, every other one is dropped."""
num = re.search(r"\d[\d.,]*", salary).group()
whole, cents = re.fullmatch(r"(.*?)(?:[.,](\d\d))?", num).groups()
period = re.search(r"\(([^)]+)\)", salary)
return (float(re.sub(r"\D", "", whole) + "." + (cents or "0")),
period.group(1) if period else None, "Comisiones" in salary)
def city_region(where):
"""'Guadalajara, Jalisco' -> two parts. Bogotá has a comma of its own: 'Bogotá, D.C., Bogotá, D.C.'."""
parts = where.split(", ")
half = len(parts) // 2
if len(parts) % 2 == 0 and parts[:half] == parts[half:]:
return ", ".join(parts[:half]), ", ".join(parts[half:])
return parts[0], ", ".join(parts[1:])
def rows_from_listing(page, country):
"""One search page -> (postings the search reports, rows). Raises ValueError on a page without results."""
head = re.search(r'<h1 class="title_page"[^>]*>\s*<span class="fwB">\s*([\d.,]+)', page)
if not head:
raise ValueError("no result count: a block page, or no results")
rows = []
for art in re.findall(r"<article class=\"box_offer.*?</article>", page, re.S):
link = re.search(r'<a class="js-o-link[^"]*" href="([^"#]+)[^"]*">(.*?)</a>', art, re.S)
place = re.search(r'<p class="fs16 fc_base mt5">\s*<span class="mr10">(.*?)</span>', art, re.S)
company = re.search(r"offer-grid-article-company-url>(.*?)</a>", art, re.S)
tags = {icon: " ".join(text(t)) for icon, t in
re.findall(r'<span class="icon (i_\w+)"></span>(.*?)</span>', art, re.S)}
city, region = city_region(" ".join(text(place.group(1)))) if place else ("", "")
salary = tags.get("i_salary")
value, period, commission = money(salary) if salary else (None, None, False)
rows.append({
"id": re.search(r"data-id='(\w+)'", art).group(1), "country": country,
"title": " ".join(text(link.group(2))), "company": " ".join(text(company.group(1))) if company else None,
"city": city, "region": region,
"salary_text": salary, "salary": value, "currency": CURRENCY[country] if salary else None,
"period": period, "commission": commission,
"work_mode": MODE.get(tags.get("i_home_office") or tags.get("i_home")),
"posted": " ".join(text(re.search(r'<p class="fs13 fc_aux mt15">(.*?)</p>', art, re.S).group(1))),
"url": f"https://{country}.computrabajo.com" + link.group(1)})
return int(re.sub(r"\D", "", head.group(1))), rows
def listing(country, title, page=1):
"""Computrabajo search for a job title in one country: /trabajo-de-<title>, ?p=2 and on for later pages."""
url = f"https://{country}.computrabajo.com/trabajo-de-{title.replace(' ', '-')}"
return rows_from_listing(get(url + (f"?p={page}" if page > 1 else "")), country)
def job_posting(page):
"""The schema.org JobPosting a posting page carries for search engines (None if it has none).
Call it with get(row["url"])."""
for block in re.findall(r'<script type="application/ld\+json"[^>]*>(.*?)</script>', page, re.S):
data = json.loads(block)
for item in data.get("@graph", [data]) if isinstance(data, dict) else data:
if isinstance(item, dict) and item.get("@type") == "JobPosting":
return item
return Nonetotal, rows = listing("mx", "contador")
print(total, "postings reported;", len(rows), "on page 1")
print({k: rows[0][k] for k in ("title", "city", "region", "salary", "currency", "period", "work_mode")})It printed 8,392 postings for contador in Mexico, 20 on page 1, and the first row: Auxiliar Contable in Venustiano Carranza, Ciudad de México DF, 12,000 MXN a month, no work mode tag. An accounting assistant is the first answer to a search for accountants, which is one of the traps below.
Read the number the way the country writes it#
Mexico writes 12,000.00, the other five countries 12.000,00. Every figure we saw had two decimals, so money() takes the last separator before two digits as the decimal point and drops the others. Read Peru's 2.500,00 with the Mexican rule and 2,500 soles become 2.5.
Do not take the currency from the sign. Four countries write $, Peru writes S/., and Ecuador writes US$ after the number, on all 60 of its salaries. CURRENCY maps the country instead; the job pages' salaryCurrency agreed with it on all 22 pages with a JobPosting.
40 of the 344 salaries end in + Comisiones, 24 of them in Ecuador. The figure is then the pay before commission, so money() keeps a commission flag.
A convenir is a 0 in the structured data#
Each posting page carries a schema.org JobPosting for search engines. On the 11 pages we opened whose search row showed a salary, it held the same number, with unitText MONTH, and the page showed the same text as the row. On the 11 whose row showed none, the page said A convenir and the JobPosting said "value": 0, MONTH. The search row already has all the pay data. If you read the JobPosting anyway, turn 0 into "not given". 2 of the 24 pages had no JobPosting at all.
Two more JobPosting fields not to trust:
validThroughwas exactly 60 days afterdatePostedon all 22 pages. That looks like a default, not a real closing date.jobLocationhas the two parts the wrong way round:addressLocalityheld the region andaddressRegionthe city, on all 17 pages where the two differ.
Clean the figures before a median#
- Same-digit figures. Argentina had 11,111 six times, 111,111 twice and 9,999 once, 9 of its 25 shown salaries. Chile had one 222,222. Drop them.
- Bulk posters. In Chile 28 of the 65 shown salaries were exactly 1,000,000 pesos, from 2 companies across 23 job titles. In the Chile vendedor search the lower quartile, the median and the upper quartile were all 1,000,000. Two companies also had 51 of Argentina's 117 postings and 51 of Chile's 120. Count postings per company and cap them, or the median is one employer's pay scale.
The search is wider than its word#
A search for contador returns accounting jobs of every kind. In Mexico 16 of the 40 contador results had contad- in the title (contador, contadora); in Peru 4 of 40. In Ecuador the programador search reported 29 postings, and only 1 had progra- in the title. Its median, USD 500, was the same as the vendedor median. Filter on the title yourself before you call a figure the pay of a job.
Which date is the age of the ad#
77 % of the rows (543 of 706) said Hace N minutos or Hace N horas. 20 of the 22 job pages with a JobPosting marked their date "(actualizada)", updated. Of the 18 whose row said hours ago, 12 had a datePosted more than a day old. The row shows when the ad was last refreshed; for its age, use datePosted.
Small things#
- Bogotá writes its place as "Bogotá, D.C., Bogotá, D.C.". Split at the first comma and the region becomes "D.C., Bogotá, D.C.";
city_region()splits it in the middle. - Pages overlap: 3 of the 40 Argentina contador rows on page 2 were on page 1 too. Keep the id and drop doubles.
- 7 of the 24 job pages listed an age range among the requirements (Edad: entre ... y ... años). It is in the requirements list on the page, not in the search row.
Every day, at scale#
For many countries on a schedule, our Computrabajo Scraper on Apify does the reading. It covers 18 Latin American countries: search by keyword and city, or paste search links. Each posting comes back with title, company with rating, city, region, salary with currency and period, work mode and the date it was posted; with "Include posting details" on, which is the default, also the contract, schedule, education, experience, skills, industry and the date it was first posted. Its workMode reads onsite also when a posting carries no remote or hybrid tag, so count only remote and hybrid. It is billed per result. How to set it up, step by step: Computrabajo jobs export.
written
11 min readWorkday jobs API in Python: a list that stops at 2,000 jobs, and dates that stop at 30+ days
Read a Workday careers site as JSON in Python: 20 jobs a request, and a list that stops at 2,000 of NVIDIA's 2,807 jobs. The two endpoints, facets that cut a big site into slices, and about forty lines of Python, measured on four career sites.
7 min readRemote job APIs: five free feeds, and the catch in each
Remote OK, We Work Remotely, Jobicy, Working Nomads and Arbeitnow publish their jobs without a key. What each feed holds, how far back it goes, the terms, and sixty lines of Python that put them in one list.
5 min readFacebook Ad Library API: which ads you get, which you don't
Meta's API returns political ads from everywhere and every kind of ad shown in the UK or the EU. An ad shown only in the US for running shoes is not in it. What each way into the Ad Library gives you, and how to keep the ads the API leaves out.