I’ve noticed that most proxy providers advertise their service mainly through price per GB, pool size, and number of supported locations.
But I’m not sure those numbers show the real cost of using a proxy for scraping.
For example, imagine two providers:
- Provider A costs $2 per GB but completes only 65% of requests successfully.
- Provider B costs $6 per GB but completes 95% successfully.
The cheaper provider may require more retries, consume more bandwidth, increase server usage, and take longer to complete the same scraping job.
At the same time, paying premium prices may not make sense for a basic website that does not have strong anti-bot protection.
So, what should users actually measure when comparing proxy providers?
My answer
I think cost per successful result is more useful than price per GB alone.
A simple calculation would be:
Total project cost ÷ number of valid results collected
The total project cost should include:
- Proxy bandwidth
- Failed requests and retries
- Server or browser-rendering costs
- CAPTCHA-solving costs
- Engineering and monitoring time
- Invalid or incomplete data
- Delays caused by unstable performance
I would also test these metrics:
- Valid-content success rate
- Average and p95 response time
- CAPTCHA and block rate
- Geographic accuracy
- Sticky-session stability
- Successful results per minute
- IP repetition rate
- Performance under higher concurrency
My current view is that cheap proxies can be the better option for lightly protected targets. Premium proxies are more likely to justify their price when retries are expensive, the target is heavily protected, precise location targeting is required, or downtime affects a production system.
I would not choose a provider based only on reputation, advertised pool size, or price. I would run a small test on the exact target and compare the real cost of obtaining usable data.
How do you evaluate proxies for your projects? Do you calculate cost per successful request, or use another metric?