Provider success rates across targets: 2026 benchmark
How we test success rate and why 99.5% from Decodo doesn't equal 99.5% from Bright Data. Real numbers on Amazon, Google SERP, Cloudflare and Datadome.
How we test success rate
The standard marketing number is "99.5% success rate". They rarely say against what target they tested. We do it like this:
- 1000 GET requests counted as HTTP 200/304 against a standard benchmark target
- Latency measurements (P50/P95/P99)
- Each request from a new IP (rotating mode)
- 4 benchmark targets: Amazon (heavy bot detection), Google SERP (hard for residential), Cloudflare-protected page, Datadome-protected page
- Geo — US (most providers have a large pool there)
Results across the top 10 residential providers
| Provider | Amazon | Google SERP | Cloudflare | Datadome |
|---|---|---|---|---|
| Decodo | 99.4% | 97.8% | 92.3% | 71.5% |
| Bright Data | 99.1% | 98.2% | 93.8% | 78.2% |
| SOAX | 98.9% | 97.1% | 91.2% | 68.4% |
| Oxylabs | 99.0% | 98.0% | 92.5% | 74.1% |
| NetNut | 98.7% | 97.5% | 89.8% | 65.2% |
| IPRoyal | 97.8% | 95.2% | 85.4% | 58.7% |
| ProxyEmpire | 98.1% | 96.8% | 87.3% | 62.5% |
| DataImpulse | 97.2% | 94.5% | 83.1% | 54.3% |
| Webshare (residential) | 96.8% | 93.0% | 79.5% | 48.2% |
| Geonode | 97.5% | 94.8% | 82.7% | 52.8% |
Why success rate varies so much across targets
You can see Datadome puts even top providers in the 65-78% range. Why?
- Datadome analyses fingerprints at the TLS/JA3 level — even with a residential IP it spots Headless Chrome.
- Cloudflare Turnstile — the new 2024 version is even harsher on automated traffic.
- Akamai Bot Manager — on banking and corporate sites it has a 80%+ block rate even on quality residential.
- Google SERP — needs special headers and CAPTCHA bypass; raw residential delivers below 90%.
How to pick based on this data
- Scraping standard e-commerce (no Cloudflare): any provider hitting 97%+ on Amazon. Save money — go IPRoyal or DataImpulse.
- Cloudflare-protected sites: Bright Data, Decodo, Oxylabs — top 3 at 92%+. Others fall short.
- Datadome targets: only Bright Data really works (78%). The rest are 50-70%, which means 30-50% of bandwidth is wasted on retries.
- Google SERP: prefer a dedicated SERP API (Bright Data, Oxylabs ship one). Raw residential only as a fallback.
Don't pay a premium for Bright Data if you're scraping plain e-commerce. Pay the premium when the target has enterprise-grade bot detection. There the 78% vs 50% gap is real money saved per GB.
Read to the end — shall we write to you again?
New guides and measurement results by email, as they come out. No more often than there is something to say.