Residential and mobile proxies are billed per gigabyte, so "how much to buy" comes before "which type to buy." Buy too much and you overpay for an unused package; buy too little and traffic runs out mid-job, forcing a redo. Below: page weight by content type, a calculation formula, and typical budgets — count before the purchase, not after.
Page weight by content type
A text results page — search, a listing, a feed without images — weighs 0.3–1 MB. A product card with photos, reviews and a recommendation block runs 2–5 MB, depending on image count and resolution. A feed with autoplaying video is a separate category: tens of megabytes per minute, and video is usually what silently drains a package budgeted on "ordinary" pages.
An order-of-magnitude gap between a text page and a video feed means content type, not proxy type, drives the final bill. Proxies priced identically per gigabyte can cost 20–30x more when the same task runs against a video platform instead of a text catalogue.
A formula for calculating traffic per task
Base calculation: number of target pages × average weight of that page type × a retry coefficient. The coefficient covers failed requests, captchas, dropped sessions and repeat calls — 1.2–1.5 for stable sites, 2+ for platforms with aggressive bot defenses. Add supporting requests separately: scripts, fonts, trackers and ad blocks that carry no target data but still count toward session traffic.
Example: 500 product cards at 3 MB each with a 1.3 coefficient is 500 × 3 MB × 1.3 ≈ 1.95 GB. Five hundred text pages with the same coefficient come to roughly 0.4 GB — a sixfold gap for the same request count, and that is the price of "heavy" content.
Typical budgets
A one-off check — opening a dozen or two pages to confirm a price or stock status — fits into tens to hundreds of megabytes and rarely needs its own calculation. A hundred product cards with images run roughly 0.3–0.5 GB per pass.
Daily monitoring of a hundred pages — competitor price tracking, ranking checks, stock checks — adds up to roughly 1.5–3 GB a month, depending on whether pages are text or media-heavy cards. A thousand text pages weigh about 1 GB, so a bulk crawl without images costs several times less than it first appears.
A large catalogue crawl — thousands of product cards with a full image set, category traversal, review collection — is measured in whole gigabytes: a thousand card checks ≈ 3–5 GB. At that scale, budget headroom upfront rather than topping up mid-task.
How to cut spend by disabling media
The biggest lever is disabling image, font and tracker loading at the automation level, not picking the cheapest channel. A browser that skips images and video still gets the same HTML and data to parse, but burns far less traffic per page. For text-only tasks — price, stock, specifications — disabling media cuts page weight 3–10x.
Saving on the channel itself — the cheapest proxy instead of a stable one — often backfires: broken sessions and retries after a dropped connection cost more than the gigabytes saved. See cheap vs expensive proxies for where that savings is illusory. An unstable channel raises the retry coefficient above, burning more traffic, not less.
Headroom for retries
A traffic calculation without headroom for retries is an optimistic scenario that rarely holds up. A captcha, a temporary IP block, a connection drop mid-page — all of it forces a retry, and a partially loaded page has already spent traffic for nothing. Budget a 20–50% reserve on top of the base calculation, depending on how aggressively the platform defends against automated traffic.
For unpredictable defenses, skip a tight fixed package and run a pay-as-you-go channel, watching burn over the first few days and adjusting the budget. Metrics worth checking on a channel before buying are in how to verify proxy quality.
Frequently Asked Questions
Which is cheaper to calculate: a residential channel or a datacenter one?
Residential and mobile traffic is billed per gigabyte, so the formula above applies directly. A datacenter address usually sells monthly at a fixed price regardless of volume — for tasks that are not geo-sensitive and involve a large request volume, it can beat any per-gigabyte calculation.
How do I check actual spend if the formula only gives an estimate?
The estimate is a starting point for buying a package, not a final figure. After the first day, compare it against actual usage in your account and adjust: if real spend runs above the estimate, the retry coefficient is likely climbing because of site defenses or channel instability.
Is traffic calculated the same way for mobile and residential proxies?
The formula is the same for both types, but mobile channels, thanks to their NAT subnet, are usually more resistant to blocks, lowering the retry coefficient on tightly defended platforms. See mobile vs residential proxies for the full difference: at equal task volume, a mobile channel can end up more economical.
Once you've sized the task, pick a channel type and set up the package in the proxy section, where current per-gigabyte rates for residential and mobile directions are listed.