Price per GB is one input to a workload budget. The useful output is the cost of a completed, validated task. The excerpts below explain what to measure, why retries matter, and how to compare two services using the same unit of useful work.
What the traffic counter measures
A common published calculation adds request headers + request body + response headers + response body. Both upload and download count. Cookies are part of the headers; they should not be added again if those headers are already included in a measurement.
Sending 2 MB to a site and receiving 1 MB back therefore involves approximately 3 MB of application data, plus any other traffic included by the billing system. A tool reporting only downloaded response bodies will show a smaller total.
The same downloaded byte passes from the website through the proxy to the customer. That does not automatically mean it is billed twice. “Upload plus download” describes the two directions of communication; the service's metering rules determine the charge.
HTTPS can add handshake and encryption-related traffic to the tunnel. Some services explicitly count those bytes. Their share matters more for repeated tiny requests using new connections than for a large download. Measuring encrypted traffic volume does not require reading its contents. Keep HTTPS protection and measure the actual cost of your connection pattern.
For browser work, the HTML is only part of the load. Images, scripts, stylesheets, fonts, background data calls, and media can all use the proxy. Extracting a few prices into a small CSV file may require downloading far more data than the final file contains.
Compression adds another distinction: transferred bytes and the size after decompression can differ. Browser network tools show both. A locally saved, decompressed response is not a substitute for the provider's traffic counter.
GB, GiB, and rounding
1 GB is 1,000,000,000 bytes. One GiB is 1,073,741,824 bytes. The same transfer has a numerical value about 7.37% larger in GB than in GiB. Commercial documentation includes an explicit change from GiB to decimal GB, so the distinction can affect real dashboard comparisons.
Check the billing unit and any minimum or rounding rule. A display rounded to two decimal places does not establish that every small request is rounded up for billing. Conversely, a headline rate per GB does not promise byte-by-byte charging under every plan.
GB also measures volume, not speed. A 10 GB package says nothing about how quickly it can be used. Connection limits and transfer rates remain separate service conditions.
Failed requests, CAPTCHAs, and paid retries
A request that fails your business task can still transfer billable data. A CAPTCHA or access-denied page has a response body even though it contains none of the information you wanted.
One published metering policy counts target-site responses such as 403, 404, and 503, plus bytes received before a transfer times out. It excludes errors generated by the proxy service itself. If the target sends nothing, the outgoing request can still count. Other products must be checked against their own rules.
A retry makes another attempt and can add more traffic. Repeatedly rotating the IP and reloading the same blocked page can consume the budget while producing no usable results. Stop and diagnose a persistent failure instead of treating every error as a reason for unlimited retries.
Error pages and successful pages may have different sizes, so a failure rate alone cannot tell you the wasted gigabytes. Measure the traffic from both. An HTTP 200 response also needs validation: a login screen, empty result, or challenge page may still be unusable.
Compare the cost of usable results
The most useful comparison is total proxy spending divided by the number of usable results. A cheaper gigabyte can cost more per completed task if the network requires many attempts.
Here is a hypothetical comparison. Each service handles 100,000 attempts, averaging 1 MB of billed traffic per attempt, for 100 GB in total:
| Service | Rate | Traffic cost | Usable results | Cost per 1,000 usable results |
|---|---|---|---|---|
| A | 2 USD/GB | 200 USD | 50,000 | 4 USD |
| B | 3 USD/GB | 300 USD | 90,000 | about 3.33 USD |
Service B costs more per GB but less per useful result in this example. In your test, use measured traffic rather than assuming every success and failure is equally large. Include expired allowance or minimum commitments when they form part of the money actually spent.
Scraping APIs can use another unit, such as successful requests or returned records, with the provider handling attempts internally. That cannot be compared with a raw proxy's price per GB until both are converted to the cost of the same completed task. An API's promise not to charge for failed requests does not automatically apply to that company's ordinary proxy network.
When unused gigabytes expire
An allowance can expire at the end of a billing cycle, last for a specified period after purchase, or remain available without a fixed expiry. These conditions affect the amount of useful traffic the payment buys.
Published arrangements include monthly GB that do not roll over, transfers of unused traffic when upgrading, and separately purchased traffic with a different expiry from the main allowance. One auto-top-up arrangement carries additional GB into the following billing cycle, then expires any remainder at the end of that cycle.
“Non-expiring” also needs its account conditions read. A published residential policy says that maintaining an available balance may depend on regular top-ups or an active subscription. Check the exact rule for your account rather than assuming the wording promises access indefinitely without conditions.
Suppose a 100 GB monthly package costs 200 USD. You use 20 GB and lose the remaining 80 GB at expiry. The advertised rate was 2 USD/GB, but the payment per used gigabyte was 10 USD. A smaller package with a higher unit price could have been cheaper overall.
If the remaining GB stay valid and will be used later, they are not already wasted. You have paid in advance for future capacity. Also distinguish canceling the next renewal from losing access immediately: some subscriptions allow the remaining traffic to be used until the paid period ends.
Reduce traffic without breaking the task
Load only what the task needs. A simple HTTP request may be sufficient for a page whose required data is already in the response. A browser is appropriate when the workflow depends on rendering or interaction. Test that distinction rather than automatically loading every page in a full browser.
Stop unnecessary resources before they download. Browser automation can block images, media, fonts, or selected third-party requests. Check that the changes preserve the required content and do not trigger extra reloads or failed attempts. Disabling JavaScript blindly can remove the data you came to collect.
Reuse work where freshness allows. Cache reusable resources, avoid duplicate jobs, and keep suitable connections or sessions alive. Repeatedly starting a fresh browser can discard reusable state and require another full load. A sticky IP may help a session, but it does not make subsequent traffic free.
Control retries and background activity. Limit attempts, pause appropriately after rate limits, and stop on persistent access or account errors. Close unused profiles and ensure system-wide proxy settings are not carrying unrelated software downloads. Keep compression enabled where supported and compare transferred sizes rather than decompressed output.
Measure the result after each change: the aim is fewer paid bytes per completed task. A large reduction in downloaded data is not a saving if the task starts failing and needs to run again.
What to establish before buying
The purchase should answer five practical questions:
- What will one usable result cost? Measure your workflow with the required network and filters, including errors and retries.
- What is included in the meter? Confirm upload, download, overhead, units, rounding, and treatment of failures.
- How much must be paid upfront? Check minimum purchases, discounts, optional features, and the price of extra GB.
- How long can the traffic be used? Read expiry, rollover, cancellation, and account-activity conditions for each allowance.
- What stops further spending? Confirm the behavior at zero balance, automatic purchases, sub-user limits, and any delay in enforcement.
With those answers, the price per GB becomes a usable budget figure: you know what is counted, what the rate covers, and how much work the purchase is likely to complete.
Adapted from ProxyData: Per GB Pricing. The selected technical sections reproduce the original guide; the introduction and selection were prepared with AI assistance.

Top comments (0)