Cheap residential proxies: cost per accepted row, not sticker price

The app counter lies; the provider meter does not

Your scraper counts what it accepted. The provider meter counts what the proxy transferred: redirect chains, challenge pages, retry bodies, images you forgot to block, and responses your parser discarded. That gap is why the app counter usually looks cleaner than the bandwidth bill.

The only reliable unit is cost per accepted row — status 200, parsed value present, correct country echo, SKU matched. Everything else is an intermediate figure.

small target test, 420 product URLs
accepted row = status 200 + parsed price + same SKU + country echo ok

fields to record:
  provider
  plan
  target host
  requests sent
  accepted rows
  provider bytes (from dashboard usage log)
  manual rejects
  policy block y/n
  notes from debugger

Proxynade Volume vs Premium is a byte-cost decision

Proxynade publishes both tiers without a sales call: Volume Residential at $0.89/GB and Premium Residential at $5.00/GB. The dashboard's Usage Logs export as CSV and show host, outcome, latency, and byte totals per request — enough to run the same target on both tiers and decide whether the premium pool earns the 5x spend. Sometimes it does. Often the cheaper pool is sufficient once images, tracking domains, and retry loops are blocked.

The expanded username carries the routing choice. A Volume Residential request with US country targeting and a 30-minute sticky window looks like rt97db6958d9-plan-volume-country-us-lifetime-30. Premium uses plan-premium in the same position. Datacenter skips the lifetime token entirely.

Webshare and IPRoyal fit different job shapes

Webshare belongs in the cheap-proxy shortlist when the target is straightforward and a monthly prepaid tier matches the job size. The important split is rotating residential versus static residential: they answer different problems, and a combined score across both hides which one actually worked.

IPRoyal's residential entry price is listed on their pricing page. The same caveat applies — run your target before committing volume. Country fit, session behavior, and retry rate vary by target, not by marketing copy.

Oxylabs and Bright Data are not cheap for small buyers

Both providers publish tiered residential pricing. Oxylabs has clear public tiers and targeting options; check their category restrictions before running a job near sensitive verticals. Bright Data has a larger published network and granular controls, but their buying path includes KYC and a policy review. If your use case is near gaming, ticketing, or account automation, confirm it passes their review before treating any number on their pricing page as final.

Neither belongs on a "cheapest" list for sub-$100 tests. They show up in comparison tables because they are well-known, not because they lead on per-GB cost for small runs.

Run the test before buying volume

The test that actually answers the question: 100 to 1,000 real requests, target host unchanged, provider-metered bytes recorded from the dashboard, accepted rows manually sampled, result written as cost per accepted row. Run it on at least two providers. The cheaper one on paper is not always the cheaper one in the CSV.