Proxy products are sold by three different units, and a headline price in one cannot be compared with a headline price in another. Each unit moves a different risk onto you.
Per GB: you carry page-weight risk
Residential and mobile products are usually metered by traffic. Cost scales with the weight of what you fetch, including pages that fail. A target that serves a few megabytes of scripts and images makes the same request ten times more expensive than a lean one. Block what you do not need, and measure bytes per successful page.
Per IP: you carry utilisation risk
ISP and datacenter proxies are typically rented per address for a term. The cost is predictable, but an idle address costs the same as a busy one, and a block of addresses on a target that blocks that range is paid for and unusable. Test a small block on the real target first.
Per request: you carry price-per-success risk
API products charge per request. The provider takes the retry and routing work, so check two things: whether failed attempts are billed, and what the definition of a successful request is. Cost per successful page is the comparable figure.
Putting them on one axis
- Per GB: price per GB multiplied by bytes per successful page, divided by success rate.
- Per IP: price per IP per term divided by successful pages that IP delivers in the term.
- Per request: price per request divided by success rate.
Every figure on the right side of those formulas is something you can measure. None of them is something a pricing page can tell you.