Guide

Pay-As-You-Go Proxies: Why a Balance That Never Expires Beats a Plan

Monthly proxy plans bill for traffic you may never send. What expiry really costs, and when a plan still wins.

HProxy Team··11 min read
HProxy.Guide

Free proxies won't hold up here.

Shared datacenter IPs get flagged and dropped fast. When it has to hold, gaming, streaming, accounts, you need mobile and residential IPs that read as a real device, from $0.44/GB, pay as you go.

See plans & pricing

A customer sent us a spreadsheet last spring. He had bought a hundred-gigabyte residential plan for a price-tracking job, run it for a month, and used thirty-four gigabytes. The scrape worked. The data was good. On renewal day the other sixty-six gigabytes stopped existing, and the rate he had compared on the pricing page turned out to be about a third of what he really paid per gigabyte of useful traffic.

Nothing about that is a scam. Every line of it sat in the terms he agreed to. It is what happens when a metered product is sold on a calendar, and it is the most expensive detail on a proxy pricing page that almost nobody checks before buying. Our how much do proxies cost piece covers the gap between an advertised rate and the rate you get on gigabyte one. This is the other gap: between the rate you get and the rate you keep.

The three shapes a proxy bill comes in

Underneath the tier tables and the plan names there are only three billing models in this market, and they behave very differently when your usage moves.

A metered subscription. You pay monthly for an allowance of traffic. ProxyRack's residential entry, for instance, is $1.10/GB on a 100 GB/month plan ($110.00/month), and Froxy's smallest residential plan is $5.50/mo. The rate looks like a per-gigabyte price. The commitment is a calendar.

A pay-as-you-go balance. You buy traffic at a published rate and spend it whenever. There is no billing period, so there is nothing for an allowance to be scoped to. This is our model, and a handful of competitors run some version of it.

A per-IP rental. You rent specific addresses for a period and traffic through them is unmetered. ISP and datacenter products usually work this way, which is why they are quoted per IP per month rather than per gigabyte. Expiry here is honest and unavoidable: you rented an address for a month, and after the month you no longer have it. Our ISP explainer covers when that trade is worth making.

The confusion lives almost entirely inside the first model, because a metered subscription wears the costume of the second. It quotes a per-gigabyte number, which is the unit of the thing you consume, while billing on a period, which is not.

The arithmetic a plan hides

An advertised per-gigabyte rate on a monthly plan is a conditional statement. It reads like a price. It behaves like a price only if you consume the entire allowance. Below that, your real rate climbs a curve, and the curve is steeper than most buyers expect.

Take that hundred-gigabyte plan at $1.10/GB, billed as a 100 GB/month plan ($110.00/month). The monthly charge does not move with usage, so the rate you pay is the charge divided by the traffic you genuinely sent.

What a $110 / 100 GB monthly plan really costs per gigabyte
Used all 100 GBthe advertised rate, earned in full
1.1 $/GB
Used 70 GB43% above the sticker
1.57 $/GB
Used 50 GBtwice the sticker
2.2 $/GB
Used 34 GBthe customer in the opening paragraph
3.24 $/GB
Used 20 GBfive times the sticker
5.5 $/GB
HProxy, gigabyte oneflat, no allowance to miss
0.5 $/GB
Source: ProxyRack published residential pricing (proxyrack.com) and HProxy list price. Effective rate = monthly charge divided by gigabytes actually sent.

Two things fall out of that chart, and only one of them is about us.

The first is that break-even sits high. A plan has to be roughly two thirds consumed before its discounted rate beats a flat rate more than double it on paper. Most buyers do not know their own usage to within two thirds, especially in the first three months of a project, which is exactly when they are asked to choose a plan.

The second is that the error is asymmetric. Overestimate your usage and you quietly pay a multiple of the sticker rate every month, with nothing on the invoice to say so. Underestimate it and you hit a wall mid-job, notice within minutes, and top up. One mistake bills you silently for a year. The other interrupts you once. Given the choice, buy the model where being wrong is loud.

The other end of the same curve

Undershooting the allowance is the expensive mistake nobody notices. Overshooting it is the one that shows up as a surprise, and the two are priced by the same design.

When a plan runs out mid-month, providers do one of three things, and which one you get is worth knowing before it happens at two in the morning. Some cut you off, which stops the scrape but at least stops the meter. Some auto-upgrade you to the next tier, which is usually the cheapest outcome and the one you least expected to agree to. Some bill overage at a rate above the plan's per-gigabyte price, on the theory that unplanned traffic is worth more, and that rate can sit meaningfully above what a flat pay-as-you-go vendor would have charged for the same gigabyte.

So the plan model can bill you above the market on both sides of its own estimate. Under-use and your effective rate climbs because you paid for traffic you never sent. Over-use and your marginal rate climbs because you sent traffic you never bought. The allowance is only the right price at exactly one point on the curve, and you are asked to pick that point in advance, before the job has told you what it needs.

A balance has no equivalent failure mode in either direction. There is no threshold to fall short of and none to exceed, because there is no allowance, only a rate and an amount you have already bought.

Why the industry likes plans anyway

Unused allowance is revenue with no cost of goods attached. Retail calls it breakage, it is the reason gift cards exist, and in a metered infrastructure business it is the highest-margin line on the sheet: the provider bought no bandwidth, paid no exit node, and kept the money. A model that profits when the customer does not use the product is not automatically dishonest, but the incentive to help you right-size your plan is missing from it.

The tier ladder pulls the same way. Discounts sit at the top of the table, so a buyer comparing rates drifts toward the largest allowance the budget allows rather than the one the workload needs, and the difference between those two is the breakage. Renewal then happens automatically, so the correction never gets a natural moment. Nothing asks you in month four whether the plan you sized in month one still matches what you are doing.

None of this requires bad faith. It is what the model optimizes for, and reading a pricing page with that in mind is a skill worth having. We wrote how to read a proxy pricing page for the wider version of the exercise.

Where a plan genuinely wins

We sell from a balance, so treat this section as the part we had to argue ourselves out of. Real cases exist where a subscription is the better purchase.

Steady, predictable, large volume. If the same scrape has run against the same target list for six months and your traffic sits inside a narrow band, you know your number. Commit to it and take the discount. Volume pricing is not a trick, it reflects what bulk bandwidth costs to source, which our supply-chain study works through in detail.

Per-IP products. When the meter is the address rather than the traffic, monthly is the natural unit. You are renting an identity for a period and unmetered traffic through it is the point. Comparing an ISP subscription against a per-gigabyte balance compares two different products.

Procurement that needs one invoice. Inside a company with a purchasing process, a fixed recurring line item is genuinely easier than a variable one, and the cost of that finance friction can exceed the breakage. Good reason, nothing to do with proxies.

A provider whose only good rate lives on a plan. If the pay-as-you-go price at a vendor you have already validated sits far above its plan rate, and your volume is solid, take the plan. Billing model is one input, not the whole decision.

Choosing between the two models

A monthly plan fits when

  • Volume is known

    six months of history inside a narrow band

  • Volume is large

    enough to reach a genuine discount tier

  • The meter is time

    per-IP products, where unmetered traffic is the point

  • Finance wants one line

    a fixed invoice beats a variable one internally

A balance fits when

  • Usage is spiky

    campaigns, launches, drops, seasonal work

  • You are still testing

    the first three months of any project

  • Work is intermittent

    an agency month with no client scrape

  • You want the error to be loud

    running out interrupts you, over-buying does not

Source: HProxy

Five questions that settle it before you pay

Every one of these has a specific answer, all five are answerable by a support agent in one message, and the answers matter more than the headline rate.

Does unused traffic roll over, and for how long? Some plans carry the remainder forward while the subscription stays active, some reset it every period, and a few carry it for a fixed number of months. Three different products, one word on the pricing page.

What happens to the remainder if I cancel? This is the question the rollover answer often hides. Traffic that rolls over "while you remain subscribed" is not yours, it is a retention mechanism, and it disappears the moment you stop paying. Ask what the balance looks like the day after cancellation.

Is the discount priced on this purchase or on a commitment? A ladder that discounts the order you are placing is simple arithmetic. A tier that discounts you for promising a monthly volume is a bet, and if you miss the volume you either lose the rate or pay for traffic you did not send. Both models are legitimate. They are not the same offer.

What is the smallest order you accept? A vendor that will sell you one gigabyte is a vendor you can evaluate for the price of a coffee. A vendor whose smallest unit is a monthly plan has told you that evaluation is not part of the relationship it wants.

Does my rate survive renewal? New-customer promotional pricing that steps up after two or three cycles is common, and it is usually disclosed in the terms rather than on the card. If the first invoice and the fourth invoice differ, you want to know that before you build a cost model on the first one.

Work out your own number

The decision reduces to one division you can do from an invoice and a usage graph.

  1. Take one month's charge. The full amount, including anything paid for overage.
  2. Take the gigabytes you actually sent that month. From the provider dashboard, not from the plan name.
  3. Divide. That is your real rate per gigabyte, and the only rate that has ever appeared on your bank statement.
  4. Repeat for three months. One quiet month proves nothing. Three months of the same gap is your billing model telling you what it is.
  5. Compare against a flat rate on your median month, not your best one. At our list price, $25 buys fifty gigabytes and $100 buys two hundred, with no date attached to either.

If the division keeps landing near the advertised rate, your plan is sized correctly and you should keep it. If it lands at double, you are paying for an allowance rather than for traffic, and the fix is either a smaller plan or a model with no allowance in it.

One second-order effect deserves naming, because it never shows up in a single month's arithmetic. A plan you are paying for regardless quietly encourages you to use it. Traffic gets spent to justify the line item rather than because a job needed it, which inflates next month's estimate, which sizes next year's plan. A balance pulls the other way: the cheapest request is the one you never sent, and that discipline tends to improve the scraper as much as the invoice.

How we do it, and the part that is not free

We sell residential from a balance at $0.50/GB for the first gigabyte, falling along a volume ladder to $0.44/GB at 2,000 GB+. The balance does not expire. No subscription exists, so there is nothing to cancel, and no renewal date decides the fate of what you have not spent. Setup is self-serve with no KYC, a separate argument we make in full in no-KYC residential proxies.

The honest catch: on the sticker, a flat rate loses to a deeply discounted bulk tier. If you move thousands of gigabytes a month on a fixed schedule, somebody's committed plan will beat our per-gigabyte number, and we would rather say so than pretend the volume discount is imaginary. Our ladder narrows that gap without closing it.

What the model buys instead is that the number on the page is the number you pay. No consumption threshold has to be reached for the rate to become true, no date forfeits the remainder, and no renewal re-prices you into a tier that stopped fitting two quarters ago. For spiky, seasonal, experimental or intermittent work, which describes most work, that is worth more than the discount.

You can check the whole claim before spending anything. Our free proxy list is live and re-checked every few minutes, and the proxy checker reports what any IP really is, ours or anyone else's. When it is time to buy, pricing shows the ladder in full, and one gigabyte is a legitimate order.

Frequently asked questions

What are pay-as-you-go proxies?
Pay as you go means you buy an amount of traffic or a number of IPs once, at a published rate, and draw it down whenever you like. There is no recurring charge, no plan to cancel, and no renewal date that decides what happens to the part you have not used. The opposite model is a monthly subscription that grants an allowance for one billing period, after which the unused remainder is gone.
Do unused proxy gigabytes expire?
On most monthly plans, yes. The allowance is scoped to the billing period, so traffic you did not send by the renewal date usually does not carry forward, and some providers only roll it over while you keep paying. On a pay-as-you-go balance like ours it does not expire at all: you bought the gigabytes, so they sit there until you spend them, whether that takes a week or a year.
Is pay as you go more expensive than a monthly proxy plan?
Per advertised gigabyte, usually yes, because a plan buys volume up front and volume earns a discount. Per gigabyte you actually use, often no. A plan's rate is only real if you consume the whole allowance, and once you drop below roughly two thirds of it, the effective rate passes what a flat pay-as-you-go price would have cost.
How do I work out my real cost per GB?
Divide what you were charged in a month by the gigabytes you genuinely sent, not by the allowance you were sold. On a plan those two numbers differ, and the gap is the part of your bill that bought nothing. Do the division for three consecutive months, because one quiet month is noise and three is a pattern.
When is a monthly proxy plan the better choice?
When your volume is steady, predictable and large enough to reach a real discount tier, and you are confident of using most of the allowance every month. Scheduled scraping against a fixed target list is the classic fit. Plans also make sense for per-IP products, where you rent a specific address for a period and the meter is time rather than traffic.
Does HProxy have a subscription?
No. You top up a balance and spend it at the published rate, the balance does not expire, and there is nothing to cancel because nothing recurs. Larger top-ups reach a better rate on the volume ladder, and that better rate applies to the purchase you are making rather than to a commitment you have to keep hitting.

Proxies that don't die mid-job

Residential, ISP, datacenter and mobile, verified by the same engine that runs tens of millions of checks. They read as a real device and hold up under load. Pay as you go, and your balance never expires. $0.44/GB is the 2,000 GB+ rate; a single gigabyte is $0.50/GB, with no minimum order.

129M+ proxy checks run · 100+ countries · HTTP / HTTPS / SOCKS · re-checked every few minutes · no signup