Residential and mobile proxy packages are mostly priced Per GB . That turns "is 5 GB enough?" straight into a budget question. Unfortunately most users underestimate their consumption badly and find their quota gone by mid-month.
In this article we look at what GB consumption really consists of, which line items get overlooked, and the measurable ways to bring it down.
What Exactly Does the Meter Count?
Providers generally count the total bytes that pass through the proxy: both request (upload) and response (download). That includes:
On a page, the HTML you actually need is often just a tenth of the total traffic. All of your savings come from managing that ratio.
Some providers count only downloaded bytes, others total traffic. If your work is upload-heavy (file transfers, large POST bodies), that distinction changes the bill noticeably. Confirm in the contract which one is counted.
A Realistic Worked Example
Say you plan to pull 10,000 product pages a day from a marketplace. The cost of three different approaches is worlds apart:
| Approach | Per page | Per day | Per month (30 days) |
|---|---|---|---|
| Full load in a browser | ~2.2 MB | 22 GB | 660 GB |
| Browser + images/media blocked | ~0.6 MB | 6 GB | 180 GB |
| Plain HTTP request (HTML only) | ~0.12 MB | 1.2 GB | 36 GB |
That is an 18x difference. You collect the same data, but the cost is completely different. That is why the first optimization is always "do I really need a browser?" is the question.
If the Network tab shows an XHR/fetch call returning the data as ready-made JSON, you no longer need to load the whole page.
If You Must Use a Browser: Resource Blocking
If you work with Playwright, Puppeteer or Selenium, blocking unnecessary resources at the network layer cuts consumption by 60–80%:
Every blocked request never reaches the proxy, so it never hits the meter either. The layout breaks, but the text content usually arrives complete.
For hands-on browser automation settings, take a look at our automation proxy page and the proxies for web scraping article.
Compression: A Free 70% Saving
Servers can send HTML, CSS and JSON compressed with gzip or Brotli — but only if the client says it wants that. Adding the Accept-Encoding: gzip, br header to your request typically saves 65–80% on text content.
Most HTTP libraries send this header by default, but it is very easy to drop by accident when you set headers manually. If you customize your header list, be sure to keep Accept-Encodingin place.
The Overlooked Line Items
Items usually left out of the math that add up to a serious share of the total:
- Redirects: Every 301/302 means an extra round trip. Long redirect chains burn both bytes and time.
- Failed requests: A request that times out still spends data. A 20% failure rate means a 20% higher bill.
- Retries: Automatic retry logic can download the same page three times.
- TLS handshakes: Every new connection adds ~5–7 KB of overhead. Using a connection pool reduces this considerably.
- Health checks: A check that runs once a minute adds up to thousands of requests a month.
- CAPTCHA pages: Blocked requests are downloaded and counted too.
A blocked request is not "free": the handshake, the block page and the retry together cost as much as two full pages. Raising the success rate is the most effective way to save.
Don't Optimize Without Measuring
Measure instead of guessing. In browser automation, summing response sizes takes only a few lines:
An average taken over a few hundred requests is accurate enough for a monthly projection. After making savings changes, repeat the same measurement and confirm the difference.
The Right Proxy Type for Each Traffic Type
If bandwidth is expensive, pushing every request through the expensive pool makes no sense. Choosing the source by traffic type lowers costs noticeably:
| Job type | Typical volume | Suitable source | Why |
|---|---|---|---|
| Static page collection | High | Datacenter | Lowest cost per GB, high speed |
| Protected marketplace | Medium | Residential | High trust score, few blocks |
| Long-session account operations | Low | ISP | Static IP, unlimited traffic options |
| The strictest platforms | Very low | Mobile | Highest trust, expensive per GB |
Static-IP packages with unlimited traffic can be far more economical than GB-based plans for high-volume, long-running work. To decide, see our ISP vs residential comparison .
Set an Alert Before the Quota Runs Out
When the quota runs out, the operation stops — which can be worse than ending up with incomplete data. A simple safety layer:
- Set a daily consumption target (monthly quota ÷ 30).
- Count downloaded bytes on your own side and raise an alert as you approach the threshold.
- At 80% switch automatically to the cheap pool; at 95% run only critical jobs.
- If the provider offers an API, sync the real quota a few times a day.
Summary
GB consumption is decided far less by "how many pages did I fetch" than by "what did I download on each page". Blocking images and tracking scripts, keeping compression on, avoiding unnecessary browser use and raising the success rate — those four together cut a typical operation's bill to a third. If you want to settle which proxy type suits your work before you start measuring consumption, take a look at our location and product pages is worth reviewing.