A proxy request uses more data than its response body. In our 6 November 2026 datacenter test, a tiny HTTPS response counted as 4,399 bytes on a fresh connection and about 191 bytes with one connection reused. These are historical measurements of one endpoint, not a fixed charge for every request.
Small responses amplify setup costs. A fresh TLS connection can move far more data than the answer itself.
Connection reuse helped in this test. The reused run counted 38,282 bytes across 200 requests, about 191 bytes per request.
Large bodies dominate. Our 1 MiB downloads counted about 1% more than the body size.
A pay-per-GB proxy measures traffic, not just the page you wanted. Understanding the difference helps you budget a crawl and decide whether connection reuse matters for your workload.
Four measured request patterns
On 6 November 2026, we sent four request patterns through the datacenter tier and read the gateway's byte counter before and after each run.
| Pattern | Response body per request | Counted per request | Counted above body size |
|---|---|---|---|
| HTTPS, fresh connection | About 13 bytes | 4,399 bytes | About 4,386 bytes |
| HTTPS, one reused connection | About 13 bytes, estimated | About 191 bytes | About 178 bytes, estimated |
| HTTP, fresh connection | About 13 bytes | 186 bytes | About 173 bytes |
| HTTPS, fresh connection | 1,048,576 bytes (1 MiB) | 1,059,234 bytes | 10,658 bytes, about 1% |
The tiny-response runs each used 200 requests. The 1 MiB run used 20. Body size is only one part of the traffic: requests, headers and connection setup also contribute.
Why tiny HTTPS responses use more data
Opening a new HTTPS connection through a proxy requires tunnel setup and TLS negotiation with the destination. The certificate chain and other handshake messages move before the response body arrives. That explains why connection setup can dominate a tiny response.
This test did not isolate each handshake component, so the difference above should not be read as a precise measurement of certificate size. Nor is 4.4 KB a universal HTTPS allowance. Another endpoint can negotiate differently.
Plain HTTP counted fewer bytes here because it did not negotiate TLS with the destination. That is a protocol comparison, not a reason to remove HTTPS from a workflow that needs encryption.
What the measured traffic would cost
The published datacenter rate checked on 14 November 2026 is $0.75 per GB. Multiplying the historical traffic by that rate gives these illustrations, using decimal GB (1,000,000,000 bytes):
| One million requests like the test | Approximate traffic | Illustrative charge |
|---|---|---|
| Tiny HTTPS response, fresh connection | 4.399 GB | $3.30 |
| Tiny HTTPS response, reused connection | 0.191 GB | $0.14 |
| 1 MiB HTTPS body, fresh connection | 1,059.234 GB | $794.43 |
These calculations combine the dated byte counts with the checked retail rate. They are not a quote for a new run. Connection reuse reduced counted traffic about 23 times for the tiny response in this test; it says little about the saving on a large page.
The current metering policy, checked 14 November 2026, counts relayed data in both directions, including headers and TLS traffic, plus a fixed 672 bytes per opened connection. Reusing a connection pays that fixed allowance once across its requests. Recheck those rules and measure a small current batch before extrapolating.
Reducing bytes without breaking the job
Reuse connections when your workflow permits it. A Python requests.Session can pool connections, subject to the server and client configuration; see the Python guide. A reused connection also keeps the same exit for that connection, so it is different from opening a new connection to request another rotating exit.
Other useful changes are to request compressed responses, avoid unneeded browser images and media, cache stable results, and use a permitted data endpoint when it supplies the information directly. For large downloads, reducing body size usually matters more than avoiding a handshake.
Common questions
Does the response body equal proxy usage? No. The public billing description includes traffic in both directions and a per-connection overhead. Headers and protocol traffic can make usage larger than the body alone; consult the current metering policy for its exact rules.
Will every tiny HTTPS request use 4,399 bytes? No. That is the rounded result of one fresh-connection pattern measured on 6 November 2026. Different destinations, connection settings and billing rules can produce different totals.
Does connection reuse guarantee the same IP forever? No. A live connection uses its selected exit, but reconnecting can select another. A sticky session requests continuity across connections; no tier guarantees a permanent IP.
Were these byte counts measured on residential proxies? No. All four overhead measurements came from the datacenter tier. They do not establish residential-tier byte counts.
Where can I check usage? The Proxies page in your dashboard shows usage and remaining balance. Compare a small job there with your own measurements before scaling it up.