Data Works Methodology: Collection, Fields and Privacy

Collected by Proxies.sx Data Works from public listing pages. Proxies.sx is not affiliated with Seminuevos, Mobiauto or Otomoto. Collected by Proxies.sx Data Works from public career boards. Proxies.sx is not affiliated with Greenhouse, Lever, Ashby or the employers listed.

The used-car collector honours each site's robots.txt. It does not request a path that robots.txt disallows, and it skips a site whose robots.txt it cannot read. The hiring collector reads the public job board endpoints that Greenhouse, Lever and Ashby document for career pages.

Used cars are collected every 4 hours per marketplace, starting on 19 October 2026. Each run has a fixed time and request budget and continues the walk where the previous run stopped, so a listing is read again only when the walk comes back to its page. Hiring boards are read once a day at 06:00 UTC. The first hiring read was on 17 October 2026, and the first scheduled daily run was at 06:00 UTC on 18 October 2026.

Seminuevos and Otomoto are walked by make and model pages, and Mobiauto by dealer stock pages. A listing is active from the first read that lists it until it is marked removed. A removal is recorded only after a complete read of the same pages no longer lists it, so a read that stops early never marks listings as removed.

The make and model walk follows a fixed order from the start of the alphabet. A sample therefore holds the makes the walk has reached, which is why the Mexico and Poland samples so far hold mostly makes that begin with A.

Prices are asking prices in local currency (MXN, BRL or PLN) as shown on the listing, not transaction prices. Listings without a price, or priced in another currency, are left out of price statistics. Quartiles use linear interpolation and are rounded to whole units.

Mileage is the odometer reading in kilometres as stated. Model years are grouped in five bands: 2009 and older, 2010 to 2014, 2015 to 2019, 2020 to 2022, and 2023 and newer.

A price change needs two reads of the same listing, and a removal needs a complete second read of the same pages. Otomoto and Mobiauto had finished one pass when these statistics were taken, and the second Seminuevos run had moved on to pages it had not read before, so both counts were zero. A zero at this stage is missing history, not a stable market.

Each run reads every board once. A job is new when its id is seen for the first time, and closed when a successful read no longer lists it. A new job is also a repost when the same board closed a job with the same title and location in the previous 60 days. Each closed job can be matched to one later job at most, and reposts stay inside the new count.

Changes are counted against the previous successful run. A board that fails to load is skipped and nothing on it is closed. A board read for the first time, including one an employer has moved to, has empty change fields until the next run.

Public pages show aggregates only. Any figure based on fewer than 10 listings is suppressed. We never publish listing ids, URLs, VINs, dealer or seller names, or individual listings. Employer names and open-job totals appear on the hiring pages because they are company-level counts from public career boards.

Row-level data, including dealer names, goes only to customers who receive the feed.

The used-car feed has 19 fields: source, listing_id, make, model, year, trim, mileage_km, price, currency, price_change, seller_type, dealer_name, city, state, vin, status, first_seen, last_seen, removed_at. Definitions are on the used-car listings page. The hiring feed has 8 fields, defined on the hiring velocity page. Each finished run downloads as CSV with these header rows.

The figures on these pages come from a dated aggregate snapshot, taken with a read-only query of the Data Works store at 19 October 2026 at 20:15 UTC. The site never queries the store. Each refresh replaces the snapshot and moves the reviewed date at the top of every page.

Sign in to Data Works in the Proxies.sx dashboard to request a free sample. The storefront needs an account, and you approve the sample and a rate before anything bills. To collect from other sites, see custom Data Works projects.