Benchmark Methodology
Residential Proxy Benchmark Methodology
Use this repeatable framework to evaluate residential proxy providers before moving scraping, SERP, browser automation, or AI data workflows into production.
What to measure before choosing a provider
A credible proxy test should measure output quality, not only raw speed. Keep target lists, concurrency, session settings, and test windows consistent so the result is useful for buying decisions.
Recommended 72-hour test plan
- 1Pick 20-50 representative URLs from your real workload.
- 2Run each provider at the same concurrency and session settings.
- 3Measure success rate, p95 latency, geo match, session continuity, and retry volume.
- 4Calculate cost per successful request using consumed traffic and any minimum commitments.
Success rate
2xx/3xx usable responses divided by total attempts
Run the same target set through each provider, count usable responses, and record retry count separately. A lower sticker price loses value quickly when retries consume traffic.
- Use the same URL list
- Run at the same concurrency
- Separate target errors from proxy errors
Latency distribution
p50, p90, and p95 response time
Average latency hides tail risk. Track p90 and p95 because long-tail delays hurt browser automation, SERP snapshots, and recurring monitoring jobs.
- Measure end-to-end request time
- Record warm and cold sessions
- Compare by target country
Geo accuracy
Country, city, and ASN match rate
A residential provider should deliver the region you asked for. Validate country, city, and ASN with an IP inspection endpoint before trusting SERP or marketplace data.
- Check country match
- Check city match where promised
- Record ASN and ISP drift
Session stability
Same-session IP continuity over a workflow
Sticky sessions should keep a stable route through multi-step workflows such as login checks, carts, forms, and browser agents. Measure continuity, not only first request success.
- Run multi-step flows
- Track IP continuity
- Record session expiry behavior
Cost per successful request
Total spend divided by usable responses
The practical cost metric is not per-GB rate. Include failed attempts, retries, idle sticky-session traffic, and monthly minimums to compare providers fairly.
- Include retry traffic
- Include minimum commitments
- Compare cost per usable output
Benchmark scorecard
| Metric | Healthy signal | Decision risk |
|---|---|---|
| Success rate | Stable usable output above your internal threshold | High retry volume raises cost and slows jobs |
| p95 latency | Predictable tail latency for your target countries | Automation and monitoring workflows become flaky |
| Geo accuracy | Country/city/ASN match the selected route | SERP and marketplace data may be polluted |
| Session stability | Sticky sessions hold IP continuity across steps | Forms, carts, and browser agents lose state |
| Cost per success | Real output cost fits your budget | Cheap GB pricing hides failed output cost |
Benchmark FAQs
Start with a proxy test, compare pricing, then review the Trust Center before scaling a workload.
How many requests are enough for a proxy benchmark?
For a buying decision, start with a small but representative set: 20-50 real URLs, repeated over multiple hours, at the concurrency you expect to use.
Should I test free proxies in the same benchmark?
Only for low-risk comparison. Free proxies can be useful for learning, but production evaluation should measure commercial providers under the same request and session settings.
What is the most important proxy metric?
Cost per successful request is usually the most practical metric because it combines success rate, retries, traffic consumed, and minimum commitments.
Run a cleaner proxy evaluation
Start with a proxy test, compare pricing, then review the Trust Center before scaling a workload.