How Proxy Pricing Works

A residential plan is sold by the gigabyte, but a scraper has no use for gigabytes. It needs pages that load with the right content. Most pricing mistakes sit in the gap between the unit on the invoice and the unit the job needs.

Five ways to bill for a proxy

Providers in the ProxyLook directory bill in five ways. The model matters as much as the price, because each one rewards a different shape of traffic.

Model	What is metered	Usually sold with	Suits	Watch for
Per GB	Traffic through the proxy	Rotating residential and mobile	Light pages, many IPs	Heavy pages and browser traffic
Per IP	Each address, per month	Dedicated datacenter, static ISP	Logins and long sessions	Traffic caps, limits on swapping IPs
Per port or thread	Concurrent connections	Rotating pools, mobile ports	Steady, bandwidth-heavy work	A fixed ceiling on parallel requests
Per request	Successful responses	Scraper APIs and unblockers	Hard targets, small teams	Surcharges for JavaScript rendering
Flat "unlimited"	Days or months of access	Unlimited-traffic plans	Very high bandwidth	Speed caps and fair-use clauses

Two jobs show the difference. A QA team keeps full browser sessions open on 30 regional versions of a site all day. That is heavy traffic over few addresses, so paying per IP wins. A crawler touches a million light pages through as many IPs as it can get. That is light traffic over many addresses, so paying per GB wins.

An "unlimited" plan always has a limit somewhere. It may be the number of threads, the speed of each connection, the size of the pool or a fair-use clause. The plan is still good value when that limit sits above what the job needs.

What list prices look like

Chapter 2 introduced the ranges below, and every calculation in this chapter starts from them. They are entry-level list prices in the ProxyLook directory at the time of writing. They are rough, they change often, and volume commitments push them lower. Live figures are on the ProxyLook pricing page.

Product	Billing unit	Rough entry-level range
Rotating residential	Per GB	Roughly $1 to $5
Rotating mobile	Per GB	Roughly $1 to $8.50
Rotating or shared datacenter	Per GB	Roughly $0.10 to $1.40
Dedicated datacenter IPs	Per IP per month	Roughly $1 to $2.50
Static ISP (static residential) IPs	Per IP per month	Roughly $1 to $3.50
Scraper APIs and unblockers	Per successful request or monthly plan	See live pages

Two things stand out. At the bottom of each range, datacenter traffic costs about a tenth of residential traffic. That gap is why the ladder in Chapter 5 starts with datacenter IPs. The ranges are also wide. The top of the residential range is five times the bottom, and a headline price does not explain the spread.

The entry price is usually the highest rate a provider charges. Larger commitments and promotions lower it, and current offers are listed on the deals page.

Cost per successful request

Price per GB and price per IP cannot be compared with each other. Neither says how often a request comes back with the right content. One number puts every plan on the same scale.

Total spendplan price plus add-ons

divided by

Successful requestsright content returned

equals

Cost per successcomparable across plans

The example below uses made-up round numbers to show the arithmetic. It is an illustration, not a benchmark of any provider. Assume every attempt transfers 500 KB whether it succeeds or fails. One gigabyte, taken as 1,000 MB, then pays for 2,000 attempts.

Plan	Price per GB	Success rate	Good pages per GB	Cost per 1,000 good pages
Plan A	$2.00	50%	1,000	$2.00
Plan B	$3.00	90%	1,800	$1.67

Plan B costs 50 percent more per gigabyte and about a sixth less per usable page. The plan with the lower price is the expensive one. It also needs nearly twice as many attempts to finish the same job, which means slower runs and more load on the target.

Real traffic is messier than the example. A block page usually weighs less than a full page, so failures cost less than successes. The only reliable inputs are measured ones: total spend and a count of good responses from a trial.

The same figure works for other models. For a per-IP plan, divide the monthly fee by the good requests that IP delivered. For a scraper API, the price is already per success, so the task is to check how the vendor defines success.

Bandwidth math

Per-GB budgets start with one multiplication: pages per month times average traffic per page, plus an allowance for retries. Take 100,000 product pages a month as an example. Fetched as raw HTML at 200 KB each, they use 20 GB. Loaded in a headless browser at 2 MB each, the same pages use 200 GB. The weights are made up for the example. The direction is not: a full browser load moves more data than a single HTML request.

The reason is simple. An HTTP client downloads one document. A browser downloads that document and then everything it references: images, style sheets, scripts, fonts, video and analytics calls. The scripts then request more.

JSON endpointLowest

Raw HTMLLow

Headless, media blockedMedium

Headless, full pageHighest

Four habits keep the meter down:

• Fetch the JSON endpoint behind a page when the page loads its data from one.
• Block images, fonts and media in the browser when the job only reads text.
• Cache anything that does not change between runs.
• Measure instead of guessing. Run 100 pages, read the usage meter, then multiply.

Many providers meter traffic in both directions, so request headers and uploads count too. Chapter 11 returns to these savings at scale. ProxyLook's proxy cost calculator turns a monthly traffic figure in GB into an estimated spend.

Hidden costs

The headline rate is rarely the whole bill. Six items deserve a line in any comparison sheet.

None of these is a trick. Each is stated somewhere in the pricing page or the terms. The work is to read them before the first payment, not after the first invoice.

Getting value from a trial

Trials come as free traffic, a low-cost starter pack or a refund window. Whatever the form, a trial exists to produce the two inputs of the formula.

1. Read the limits first. Trial accounts may cap traffic, restrict targeting or ask for identity checks. Refunds often lapse above a usage threshold.
2. Test the real target. Fetching a test page proves the credentials work and nothing else.
3. Define success in advance. Pick an element only the real page contains and check for it in every response.
4. Log four numbers. Record requests sent, successful requests, gigabytes used and typical response time.
5. Test at production shape. Use real concurrency, real locations and the hours the job will run.
6. Compute cost per 1,000 successes. Use the plan price that would apply after the trial, add-ons included.

Run the same script against two or three providers and the decision usually makes itself. ProxyLook's compare page helps build that shortlist before any money moves.

Key takeaways

• Proxies are billed per GB, per IP, per port or thread, per request, or as flat plans with limits elsewhere.
• List prices are rough and change often, so check live pricing before budgeting.
• Cost per successful request is total spend divided by responses with the right content.
• Headless browsers can use many times the traffic of raw HTML requests for the same data.
• Expiry, rollover rules, minimum top-ups and paid add-ons belong in every price comparison.
• A trial is only useful when it measures real targets at production volume.