Home / Methodology
FOLLOW THE SOURCE

Every number has a source.

How public provider pages become comparable, traceable pricing data.

Methodology

This page describes exactly how VPS Deals Wire turns public web pages into the numbers you see on this site. It is written against the site's own build, not against marketing copy — everything below can be checked in the public repository. Last reviewed: 2026-10-23.

The pipeline

The whole site is generated by a deterministic three-step pipeline. There is no runtime AI, no manual editing of prices, and no step where a number can be typed in by hand:

  1. Rules. .ilang/site.ilang declares which providers are tracked, which fields are extracted, and the quality rules below. It is the single source of truth: behavior changes start there.
  2. Scraper. Every 6 hours an automated job fetches each provider's own public pricing or promotions page and extracts structured entries. A page that does not answer HTTP 200 is marked unreachable — never worked around, never filled in from elsewhere.
  3. Build. The extracted data is rendered into static pages and deployed. Pages are rendered purely from the data plus the rules — there is no step where a number can be typed in by hand.

What is extracted per offer

Every deal record carries only these fields, each taken from the provider's page:

  • title — the offer label as published (used only internally; card titles are generated from provider name + price instead)
  • price and currency — exactly as published, never converted
  • offer_url — where the offer can be bought
  • source_url — the provider page the figures were read from
  • valid_until — when the provider says the offer ends, if stated
  • fetched_at — when this entry was last read; shown on every deal card as its verification time

Public dataset

The price data behind this site is published as an open dataset (CC BY 4.0): doi:10.5281/zenodo.23293595. Each record carries its official source page URL and the date it was last verified.

Quality rules

  • No price, no price field. If a price cannot be verified together with its currency and URL, the entry is published without a price — never with an estimate.
  • Expired offers are labelled. An offer past its valid_until is marked expired, not silently kept.
  • Respect source access rules. robots.txt is checked using our declared bot identity, including before following a pricing-page redirect. Missing robots files (404/410) permit access; denied or unavailable rules stop the fetch. Unreachable is unreachable. Providers whose pages cannot be fetched are marked as such on the comparison table. No proxy tricks, no spoofed user agents, no login-walled content.
  • Counts with a plus are lower bounds. Where a page holds more qualifying offers than get extracted, the count reads 12+ — "at least this many". A count without a plus sign is exact.
  • One exit per offer. Offers live inside their provider's page; there are no separate per-offer URLs to go stale. Old per-offer addresses redirect (301) to their provider's page.

What this site does not do

  • No payment for inclusion or ranking in the provider directory. Separate homepage affiliate placements are explicitly labeled Sponsored.
  • Some provider-page “Official site” buttons are affiliate links; they have a nearby disclosure. Source links are untagged. The homepage Sponsored section also contains affiliate links. Current providers and networks are listed in the privacy policy.
  • No exchange-rate conversion or cross-currency ranking. Amounts retain their source currency. A monthly price is identified only when the source explicitly states the billing period; older snapshots without that field are marked unverified.
  • No invented offers, prices, discounts, commissions, companies, or user counts — the repository's rules forbid each of these explicitly.

How providers get listed, who runs the site, and how to report an error are on the about page. For data handling, see the privacy policy.

Corrections

If a figure here disagrees with the provider's own page, email contact@vpsdealswire.com. Fixes go into the source rules and the scraper, so a correction holds for every future refresh — not just for one page load.

How the numbers get used