I buy proxies. I test them.
I publish what happens.

Twenty sites. Roughly 2,400 requests per provider, per run. Residential today. Mobile, ISP, datacenter, scraping APIs and SERP APIs are on the way.

20 sites · 6 providers · ~2,400 requests each · last run 2026-09-16

Latest report

September 2026: 6 providers, 20 sites, ~2,400 requests each.

Decodo and DataImpulse tied for the top. Full run: see the data →

#ProviderSuccessTimeoutp50 latencyCost / 1kCeiling
01decodo76.9%0.1%1,191 ms$0.63c8
02dataimpulse76.8%2.8%2,114 ms$0.15c16
03oxylabs72.4%0.3%1,180 ms$0.86c8
04webshare71.7%0.5%1,242 ms$0.51c8
05shifter70.1%4.5%2,402 ms$0.33c8
06byteful69.3%1.4%961 ms$0.46c8

Run: 2026-09-15 / 16 · residential proxies · sequential across providers · no retries · from probe_set.json

Latest activity

What's changed lately.

Each entry ships as a git commit. See the full changelog →

  • DataInitial 20-site probe run across 6 providers, ~2,400 requests each
  • ReviewDataImpulse review: cheapest per-request cost, tied for top success
  • ReviewDecodo review: fastest of the six with the lowest timeout rate
  • ReviewByteful review: snappiest median latency, ceiling at c8
What I test

Seven tests. Same script for every provider.

Every headline number on this site comes from one of these.

01

Concurrency

How the network holds as parallel sessions ramp: 1 → 4 → 8 → 16.

metric: max concurrency clean
02

IP rotation

How often IPs actually change across the pool.

metric: unique IPs per 1k
03

Sticky sessions

Does a sticky session actually hold the same IP over time?

metric: session hold rate
04

IP pool

Country and ASN distribution. Uniqueness and repetition.

metric: ASN + country spread
05

Geography

Does a proxy tagged as US actually egress from the US?

metric: geo match %
06

Latency

Connect + response time under load. p50 and p95.

metric: ms per request
07

Reliability

Success, block, and timeout across the full 20-site set.

metric: success / block / timeout %
Use cases

What real workloads actually need.

A proxy that works for SERP tracking is not the same proxy that works for Amazon price monitoring. One report per workload as I run each one.

SERP trackingPrice monitoringE-commerce (Amazon)JS-heavy scrapingAds monitoring
Why this exists

Because "millions of IPs" tells you nothing.

I've been using proxies for scraping and SEO work for years. Every provider says the same things: millions of IPs, fast, high success rate, rotating residential. None of those tell you what happens when you actually put eight concurrent sessions through the network on a site behind PerimeterX.

So I started testing them myself. What began as a script to check if a provider could handle my SERP scraper turned into twenty sites and roughly 2,400 requests per provider, a ranking that moves month to month. proxy.report is where I publish what I find. Read the full story →

Follow along

New reports drop when the tests land.

No newsletter yet. RSS and GitHub are enough.

The plan

What's on this site, and what's coming.

Every page here is one of four things. Here's the state right now: done in yellow, coming in gray.

01

Reviews

One essay per provider: opinions, tradeoffs, verdict.

  • Byteful
  • DataImpulse
  • Decodo
  • Oxylabs
  • Webshare
  • Shifter
02

Tests

What I measure and why it matters when you actually scrape.

  • Concurrency
  • IP rotation
  • Sticky sessions
  • Latency
  • Reliability
  • · Geo reliability
  • · Anti-bot bypass
03

Use cases

How providers hold up under real workloads.

  • · SERP tracking
  • · Price monitoring
  • · E-commerce (Amazon)
  • · JS-heavy scraping
  • · Ads monitoring
04

Best-of

Annual rankings per proxy type. The commercial pages; rankings still come from the same test data.

  • · Best residential proxies 2026
  • · Best cheapest proxies 2026
  • · Best fastest proxies 2026
FAQ

Questions I've had to answer.

Is this site paid by proxy providers? +

Some providers are affiliate partners. Some aren't. Rankings come from the same test data everything else uses. If a provider I earn from loses on the numbers, they lose on the page. Every provider page tells you whether I make money if you buy through it. See the full affiliate log.

How fresh is the data? +

I re-run the full 20-site test monthly. Every review cites the specific run it was written against. When numbers change, the review and the changelog both update.

Why only residential proxies right now? +

I started with residential because my first use case was SERP tracking. Mobile, ISP, datacenter, scraping APIs, and SERP APIs are all coming. Each provider page will grow a new report tab as the data lands.

Can I re-run these tests myself? +

The test scripts are being cleaned up for public release. Once they're on GitHub, you can run the same script against the same sites and check whatever I've written.

What providers are being tested next? +

See the roadmap. Short answer: more providers, more proxy types, and use-case-specific essays.