How Much Bandwidth Do You Actually Need?

Metered usage is measured at the proxy, in both directions. Every byte that crosses the connection is counted — not just the response body you were after.

• Request bytes — the URL, headers, cookies, and any POST body you send.
• Response bytes — headers plus body. Usage is the bytes that actually cross the connection — the transferred size, not the decoded size your client reports.
• TLS overhead — handshakes and record framing on HTTPS connections. Small per connection, but it adds up if you open a fresh connection for every request.
• Everything else — redirects, retries, 404s, and failed attempts all transfer bytes, so they all count.

The practical takeaway: your real cost per page is the sum of every request the page triggers, not the size of the HTML file alone. A single "page" fetched in a headless browser can fire dozens of sub-requests for scripts, styles, images, and trackers.

Honest ranges, measured as transferred bytes. Your targets will vary — treat these as planning numbers until you measure your own.

• JSON API response: 1–50 KB. The cheapest way to move structured data.
• Plain HTML page (no assets): 50–200 KB. What you get with a bare HTTP client that fetches only the document.
• Modern JS-heavy page, fully loaded: 1–3 MB or more. Frameworks, fonts, images, ads, analytics — a headless browser loading everything pays for all of it.
• A single image: 100 KB–1 MB depending on dimensions and format.
• Video: tens to hundreds of MB per clip.

Using 1 GB ≈ 1,000,000 KB, the math is simple multiplication. Three jobs at realistic weights:

# Lightweight product pages (HTML only)
100,000 pages x 150 KB = 15,000,000 KB ~= 15 GB

# API polling
1,000,000 calls x 20 KB = 20,000,000 KB ~= 20 GB

# JS-heavy pages in a headless browser, assets on
10,000 pages x 2.5 MB = 25,000 MB ~= 25 GB

Notice the spread: the browser-rendered job fetches 10x fewer pages than the HTML job yet costs more. Same target category, wildly different bill — the difference is entirely in what you fetch, which is why the next section matters more than any pricing detail.

Most metered-proxy bills are dominated by bytes nobody looks at. These changes are cheap to implement and compound:

• Request only what you need. If the data is in the HTML, use a plain HTTP client and skip the browser entirely.
• Prefer official APIs or JSON endpoints where the target permits their use — a 20 KB response instead of a 2 MB page.
• Disable images, media, and fonts in headless browsers. Blocking those resource types typically cuts a rendered page's transfer by 50–80%.
• Send `Accept-Encoding: gzip, br` so responses arrive compressed. Most HTTP libraries do this by default — verify yours does.
• Cache and dedupe. Never fetch the same URL twice in one run; normalize URLs so tracking parameters don't create duplicates.
• Skip unchanged pages. Use ETag / Last-Modified conditional requests, or sitemap timestamps, so re-crawls only pay for what actually changed.

Don't estimate from page-weight tables alone — measure your own workload. A small pilot gives you a real per-request number in minutes.

1. 01Buy a small amount first — the 10 GB quick plan or any custom amount from the 1 GB minimum up covers most pilots. See how to buy GB.
2. 02Run a representative sample: around 1,000 requests against your actual targets with your actual client configuration.
3. 03Check consumed data in the dashboard usage stats and divide by the request count to get your true cost per request.
4. 04Multiply by your total planned volume, then add a 20–30% buffer for retries, redirects, and growth.
5. 05Size the real purchase to that number. The Plans page previews the live discounted price before you commit — discounts reach up to 30% off at the largest tiers.

If your monthly arithmetic keeps climbing — sustained heavy crawling, large media-adjacent workloads, or throughput you can't meaningfully reduce — flat-rate unlimited plans can beat per-GB pricing. They use separate credentials from your GB plan and come in three speed tiers: 200 Mbps (hourly only), and 400 Mbps or 1 Gbps in daily, weekly, or monthly terms.

The tradeoffs: capacity is slot-limited per tier and can sell out, and you're paying for a time window rather than a byte count — idle hours are wasted money in a way idle GB never are. Run the comparison yourself: take your measured monthly consumption at your effective per-GB price and set it against the current rates on the Unlimited Proxies page.

A reasonable default for most users: stay on pay-per-GB while your usage is spiky or unpredictable, trim waste first, and move to unlimited only once your measured baseline makes the flat rate the obviously cheaper column.