Pay-As-You-Go vs Subscription Proxies: Cost Guide

Two proxy buyers with identical traffic can pay wildly different amounts — not because one found a secret discount, but because one picked the wrong billing model. The choice comes down to pay-as-you-go (per-GB, no commitment) versus a subscription (a monthly plan with an included quota). Neither is universally cheaper. The right answer depends on how steady your volume is and how predictable it stays.

With o2co you get both off the same proxy endpoint, so switching models is a billing decision, not a migration. This guide explains how each model works here and how to choose well.

How per-GB pricing actually works

Pay-as-you-go bills the bytes that pass through the proxy, in both directions, with no monthly minimum. On our endpoint you top up credit (currently in USDT) and either spend it as you go at the per-GB rate or turn it into a data balance by buying GB up front. Neither expires — there's no “use it or lose it” and nothing to cancel. If you send no traffic this month, you pay nothing.

We meter a gigabyte as 1,000,000,000 bytes, and the meter is the same whether a connection lands on a residential or a datacenter exit — one per-GB rate, shown on the pricing page.

What a billed “GB” really measures

This is where buyers underestimate their bill. A billed gigabyte is not the size of the HTML you wanted — it's the total traffic on the wire to deliver it. That includes:

• Request and response headers on every call, which add up fast across thousands of small requests.
• Sub-resources — a single page-load pulls in scripts, CSS, images, and fonts. The page you see is a fraction of the bytes transferred.
• Retries and failures. A request that returns a block page or dies mid-download still moved bytes, and you pay for them. (A connection that never opens moves nothing.) Noisy retry logic can multiply your usage.
• TLS and protocol overhead — handshakes and framing on top of your payload.

Rule of thumb: budget meaningfully more than your “useful” payload. If your scraper logically needs 10 GB of content, plan for noticeably more once headers, sub-resources, and retries are counted. Tightening retries and skipping resources you don't need (images, analytics beacons) is the cheapest optimization there is — we cover that pacing in Proxies for Web Scraping.

When pay-as-you-go is cheaper

• Low or occasional volume. If you use a few GB a month, most of a plan's included quota would expire unused.
• Spiky, unpredictable traffic. Campaigns, one-off research, seasonal jobs — you don't want to size a subscription to a peak you only hit twice a year.
• Testing and evaluation. Proving out a target, comparing residential vs datacenter, or validating your setup shouldn't require a commitment. Pay for exactly what the test burns.
• Multiple small clients or projects. Metering each one separately is cleaner than carving up a fixed quota.

When a subscription wins

• Steady, high volume. A plan includes a monthly GB quota at a lower effective per-GB rate than pay-as-you-go — worth it once you reliably use most of that quota. Committed volume is what earns the discount.
• Predictable budgeting. A flat monthly line item is easier to forecast and approve than a balance that moves with usage.
• Always-on workloads. Monitoring, continuous verification, or production traffic that never really stops fits a quota model naturally.

How the two combine here

On o2co a plan and a data balance stack. Each byte is drawn in a fixed order:

1. Plan quota — the GB included in your plan for the current period. The period runs monthly from the day you bought the plan, and unused quota expires at the end of it.
2. Data balance — GB you bought separately (or were granted at signup). It never expires and is only touched once the plan quota is used up.
3. Credit — if both are empty, traffic continues from your prepaid credit at the plan's overage rate. Traffic stops only when all three are exhausted.

Rough heuristic: start on pay-as-you-go. Watch a few months of real usage. If your monthly GB is consistently high and stable, move the baseline onto a plan — and keep a data balance for the spikes on top, since it waits for you instead of expiring.

Why residential usually costs more per GB

Across the market, the biggest driver of the per-GB rate is the type of egress IP. Datacenter bandwidth is cheap and abundant; residential bandwidth rides real ISP connections that cost far more to source, so most providers price it well above datacenter, and some add a surcharge per country. We don't split the rate that way: one per-GB rate covers whichever exit and country a connection lands on. Why targets treat the two types differently is in Residential vs Datacenter Proxies.

When comparing residential offers, ask how the supply is sourced. Bandwidth taken from devices whose owners never knowingly agreed to share it is an ethical problem and a reliability one, because that kind of pool can be cut off without warning. An unusually low price is a reason to ask that question, not proof of anything by itself.

Avoiding bill shock

Per-GB billing's flexibility is also its risk: a runaway script can burn a balance fast. Protect yourself before you scale up.

• Prepaid is its own cap. Nothing here is billed in arrears and there is no auto-top-up: when your quota, balance, and credit run out, traffic stops. Size the top-up to what you're willing to spend.
• Cap retries and add backoff. Infinite retry loops on a blocked target are pure wasted bytes.
• Fetch only what you need. Skip images, media, and third-party beacons when you're after structured data.
• Watch the first run closely. Measure a small batch, extrapolate, then turn up the volume.

Side by side

It's the same endpoint either way

Whichever model you're on, the connection doesn't change: one proxy host (the one on your dashboard) that speaks HTTP and SOCKS5 on the same port, with username/password auth. You reach residential and datacenter exits through it, and you target a country by appending a lowercase -country-xx to your username (for example USERNAME-country-de to exit through Germany). Billing model is something you buy in the portal — not a different product, not a different integration.

# same endpoint, PAYG or plan
curl --proxy http://USERNAME:PASSWORD@HOST:PORT https://api.ipify.org

# route this connection through Germany
curl --proxy http://USERNAME-country-de:PASSWORD@HOST:PORT https://api.ipify.org

# prefer SOCKS5? same port, same credentials
curl --proxy socks5h://USERNAME:PASSWORD@HOST:PORT https://api.ipify.org

The practical upshot: pick pay-as-you-go to start and to handle spikes, graduate your steady baseline onto a plan when the numbers say so, and size your top-ups deliberately. You're optimizing a billing setting, not rebuilding your stack. Get 1 GB free (credited as a data balance that never expires) or compare current plans and rates on the pricing page.