Summary of an article published on The Web Scraping Club on 2025-04-29 (free to read). Read the full article

Residential proxy providers advertise pools of 70,000,000+ IPs, and plenty of buyers assume more is better. Those figures usually count every unique address that has passed through the network over the years, not what is active and usable for you today. For most scraping projects, raw IP count is the wrong thing to optimize. IP quality, concurrency and how the scraper behaves matter more.

We work out how many proxies a project needs from three inputs: the request volume you must hit, how much the target site tolerates per IP, and a safety buffer. Then we apply that to Amazon at four scales, from testing a single product page up to enterprise volumes, with setup, risk level and monthly cost for each. We also look at why sites actually block you, and close with the questions to put to a provider before paying.

What’s inside

  • Why lifetime IP totals say nothing about the IPs available to you right now, with a rideshare analogy.
  • Per-IP tolerance differs by site type, and we compare social media, e-commerce and news sites.
  • Blocks come from behavior, headers and datacenter IPs, and a bad pattern can taint a whole subnet. Subnet reputation is left for a future post.
  • Questions for providers: daily active IPs in your target regions, subnet diversity, ethical sourcing, session management, performance on your sites and trial terms.

In the full article: Proxy counts, rotation settings and monthly cost ranges for each Amazon scenario, the sizing formula with its worked example, and the complete list of provider questions.

Read the full article on The Web Scraping Club