Long-form teardowns for teams that have to defend a scraping bill. Every report starts from the invoice, normalizes it to cost per 1,000 usable pages, and shows the arithmetic you can re-run on your own targets.
A 100,000-request crawl can bill 160,000 attempts once retries enter the picture. This teardown separates the sticker price from the usable-page price across ZenRows, ScrapingBee, ScraperAPI and Crawlbase, then gives you a two-hour pilot that produces a buying decision instead of a vendor ranking.
A reseller quotes $0.49 per 1,000 requests. At 70% usable yield the real price is $0.70, and a naive three-retry policy pushes it to $1.12. The check-proxy-tax CLI measures status codes, blocked reasons, latency percentiles and effective cost per 1,000 usable responses with no account and no dependencies.
How to read these numbers. The managed-provider figures are supplied planning baselines observed on 2026-10-17, not independent target measurements or current vendor quotes. The CLI results were produced by local runs on 2026-10-19. A 70% yield figure is not a guarantee for your targets: run the probe against your own URL list and marker before you commit to a plan.
Clone the zero-dependency CLI, point it at the hardest URLs in your list, and get cost per 1,000 usable responses in JSON. Then compare that number with the live benchmark.
Get check-proxy-tax ↗Open the live benchmark ↗Calculate true TCO ↗