The meter bills what crossed the proxy, not what the parser kept
A pool does not become cheap until you check accepted rows. Volume Residential at $0.89/GB is the right buy on many targets. On others it is a slow way to pay for 403s, blank bodies, and retries. The per-GB bill counts all of it.
The log row that changes a plan decision is straightforward: route name, target URL, status code, provider-metered bytes, saved rows, and the exact proxy line used. Without it, every pricing argument is guesswork.
Here is the kind of export that makes me stop a run instead of adding threads. Host name redacted, numbers rounded.
host: retail-search-redacted
plan: volume residential
price: $0.89/GB
billed: 9.43 GB
accepted rows: 3,118
blank bodies: 1,904
403 or challenge page: 2,771
retry bodies kept by nobody: 684
rough cost: $3.02 per 1,000 accepted rows
same script, premium test
price: $5.00/GB
billed: 1.76 GB
accepted rows: 7,420
429 burst: 181
bad bodies: 96
rough cost: $1.19 per 1,000 accepted rowsThe cheap pool did not fail every request. It failed enough of them, and pushed enough junk through the browser, that the $0.89/GB plan stopped being cheaper than the $5.00/GB test. That is when low-cost rotating proxies hurt results.
The same arithmetic applies to datacenter when the target is sensitive to ASN reputation or rate shape — blocklists like Spamhaus track datacenter ranges that repeatedly trigger MDN: 403 Forbidden at scale. Public product pages may stay clean. Checkout-adjacent flows, logged-in dashboards, and regulated data portals give less room for noisy routing. The failure mode is not always a clean block — often it is a growing pile of empty responses and retries that look harmless until the bill lands.
App bytes versus proxy bytes
Apps lie about bandwidth because they show what the app kept, not what crossed the proxy. A scraper may save one parsed row while the browser has already pulled redirects, scripts, images, TLS setup, background calls, challenge pages, and failed attempts through the pool.
Check provider-metered GB against accepted output rows. If the app reports a small run but the Proxynade usage log exports several GB, the app number is not the one that gets billed. Blocks and retries are not free just because the parser discarded them. The dashboard network logs show host, outcome, latency, and byte totals; the usage export is the authoritative billing view.
When to test up to Premium Residential
Datacenter is the right start for targets that do not scrutinize ASN. Volume Residential at $0.89/GB holds up when the success rate stays healthy. Premium Residential at $5.00/GB earns a test when the cheaper pool is buying bad bodies faster than good rows. Static ISP is a different conversation — it makes sense when the workflow carries account state, fixed geography, or a saved browser profile, not as a substitute for a rotating pool.
The only calculation worth trusting is cost per successful result after retries, blocks, empty responses, and bandwidth waste are included. Sticker price is just where the argument starts.