Skip to content
LightningBytes
Back to Blog

What Is Proxy Bandwidth and How Much Do You Need?

Proxy bandwidth is the metered unit you pay for. Here is how to estimate your needs from real request sizes, and how to cut waste before you buy.

by LightningBytes Team
  • proxy-basics
  • proxy-management

Most residential and mobile proxy plans are priced per gigabyte, which makes bandwidth the number you are actually buying. Estimating it badly means either running out mid-project or paying for headroom you never touch.

This is how to size a collection job, and how to reduce the bytes you spend on the way.

What counts as proxy bandwidth

Bandwidth is the total volume of traffic that passes through the proxy, counted in both directions. When you fetch a page, the bytes that count include:

  • The request itself. Small, but not zero, especially with large headers and cookies.
  • The response body. The HTML, JSON or binary content.
  • Sub-resources. CSS, JavaScript, images, fonts and analytics requests, if your client loads them.
  • Retries. Every retry is another full round trip.

The last two are where estimates go wrong. A "small" page can be several megabytes once a browser loads its full asset set.

Estimating your needs

A rough method that gets you close:

  1. Sample the target. Fetch a representative page through the proxy and measure the transferred bytes. Do it for several page types, because a product page and a search results page differ.
  2. Count requests per run. Include pagination, detail pages and any API calls.
  3. Add retries. If your failure rate is 10 percent, add roughly that proportion again.
  4. Multiply and convert. Requests multiplied by average bytes gives total volume.
  5. Add headroom. A margin of 25 to 50 percent covers spikes, schema changes and debugging.

Then check how your figure sits against package tiers, since stepping just over a boundary often lowers the effective unit rate. That trade-off is covered in How Much Do Proxy Services Cost.

A worked estimate

Suppose a monitoring job covers 5,000 product pages, four times a day.

  • Average page with assets: about 1.5 megabytes.
  • 5,000 pages times 4 runs is 20,000 requests per day.
  • 20,000 times 1.5 megabytes is 30 gigabytes per day.
  • A 10 percent retry rate brings it to about 33 gigabytes.

That is a meaningful daily spend, and it is dominated by assets rather than by the data you wanted. Blocking them is the first optimisation.

Cutting usage without losing data

These changes routinely reduce bandwidth by a large factor:

Block non-essential resources. If you only need text, do not load images, video, fonts or tracking scripts. Browser automation makes this easy with request interception, and the savings are often most of the total.

Find the underlying API. Many pages render from a JSON endpoint that is a fraction of the size and far easier to parse. The method is described in Stop Scraping the Page, Find the API Instead.

Cache aggressively. If a page has not changed, do not refetch it. Conditional requests with If-Modified-Since or ETag can turn a full download into a few hundred bytes.

Cap response size. Abort responses that exceed an expected size, which also protects against accidentally downloading large binaries.

Reuse connections. HTTP keep-alive avoids repeated handshakes, which matters more for many small requests than for large ones.

Fix your selectors. Parsing the same page twice because the first attempt failed costs full bandwidth again. Correct selectors are a bandwidth decision.

When bandwidth is not the constraint

Some workloads are request-bound rather than bandwidth-bound. If you are making thousands of tiny API calls, the per-gigabyte meter barely moves, and your limits come from concurrency and rate limiting instead. In that case, evaluate providers on session control and request allowances, not on gigabytes.

The mirror case is media-heavy collection, where a small number of requests consume enormous bandwidth. There, the per-gigabyte rate dominates everything.

Measuring rather than guessing

The only reliable estimate comes from your own traffic. Run a small sample, log bytes per request, and extrapolate. If you are unsure whether an endpoint is behaving, the proxy checker confirms the exit IP and latency, and IP lookup verifies the geography.

For sizing guidance around a first purchase, How Much Residential Bandwidth Do You Need walks through the same method with residential-specific notes, and the residential page lists current package sizes.

Start working with cleaner IPs

Clean, pre-filtered residential and mobile proxies, sign up and send your first request in minutes.

We use cookies for authentication and security. With your consent we also enable optional marketing & analytics cookies. See our privacy policy.