Unmetered vs Pay-Per-GB Proxies: Real Cost Breakdown

Two Pricing Models, One Question

Residential proxy providers charge you in one of two ways. Per-GB pricing bills you for every gigabyte of data that passes through the proxy. You pay based on consumption. Unmetered pricing charges a flat monthly rate regardless of how much bandwidth you use. You pay based on concurrency (threads), not traffic volume.

Both models work. Neither is inherently better. The question isn't which pricing structure exists. It's which one costs less for your specific workload.

If you're running light, occasional tasks, per-GB can be the cheaper option. But most people reading this aren't running light, occasional tasks. You're running scrapers at scale, managing dozens or hundreds of social accounts, monitoring competitor prices around the clock, or automating workflows that never stop. And that's where the math gets interesting.

The problem is that most people underestimate their actual bandwidth consumption by a wide margin. They look at the per-GB rate, estimate their data needs based on a best-case scenario, and end up shocked when the invoice arrives.

Let's look at what your operations actually consume, because it's almost certainly more than you think.

What Burns Bandwidth (That You Don't Expect)

When people estimate their bandwidth needs, they usually calculate based on successful response sizes. A product page is 200KB. I need a million of them. That's roughly 200GB. Simple.

Except it's never that simple. Here's what they miss.

Retries on failed requests. When a request fails (timeout, block, rate limit), your scraper retries it. On per-GB plans, every retry burns bandwidth. Even the failed ones. A 30% failure rate means you're paying for 30% more data that gave you nothing useful.

Redirects. A single URL can chain through 2-3 redirects before reaching the final page. Each redirect is a separate request with its own headers and response body. That 200KB page might cost 350KB in total transfer.

CAPTCHAs and challenge pages. When a site serves a CAPTCHA instead of your target page, you still download the entire challenge payload. These HTML pages with embedded JavaScript can run 500KB or more. You pay for data that contains zero useful content.

Headers, cookies, and TLS overhead. HTTP headers, authentication cookies, and encryption overhead add 2-5KB per request. On a million requests, that's 2-5GB of pure overhead before you've downloaded a single page.

The result: your actual bandwidth usage is typically 3-5x what you'd estimate from successful response sizes alone.

That 3x multiplier is the number that wrecks per-GB budgets. You plan for $500/month and get invoiced for $1,500.

The Math: Three Real Workloads

Let's run the numbers on three common use cases. These estimates use conservative, real-world bandwidth figures that include retries, redirects, and overhead. Per-GB rates are based on typical residential proxy pricing: $10/GB for committed plans and $15-20/GB for pay-as-you-go.

Web Scraping (1M requests/month)

A million requests per month is mid-range for most scraping operations. Product pages, SERPs, social profiles, real estate listings. You need consistent throughput and can't afford to throttle your pipeline because of billing anxiety.

At the low end, you save $500/month. At the high end, you save over $3,000/month. The thread count you need depends on concurrency, not bandwidth. 100 threads (Starter at $79/month) handles most scraping operations comfortably. Heavy operations with parallel requests across thousands of pages, or jobs that need to complete within tight time windows, might need 500 threads (Growth at $249/month) or 1,000 threads (Scale at $499/month). Either way, bandwidth is unlimited on every plan.

Social Media Management (200 accounts, 24/7)

Running social accounts means persistent connections. Each account loads feeds, renders media previews, posts content, and interacts with other profiles. That's steady bandwidth consumption across every active session.

Social media bandwidth is harder to predict because it depends on content type. Image-heavy feeds burn more data than text-based interactions. Video previews add up fast. And the platforms themselves keep changing how much data each session consumes as they roll out new features and heavier page loads.

With per-GB pricing, a viral content day that drives extra feed refreshes across your accounts can spike your bill unexpectedly. With unmetered, you don't think about it. Your accounts run. Your costs stay the same.

Price Monitoring (10,000 SKUs, hourly checks)

E-commerce price monitoring means hitting the same pages repeatedly, often hourly. Product pages, category pages, search results. Each check downloads the full page with images, JavaScript, and dynamic content.

Price monitoring is where per-GB pricing hurts the most. The combination of high frequency (hourly), high volume (10,000+ SKUs), and heavy pages (modern e-commerce sites love JavaScript) creates bandwidth usage that balloons fast. This is also where retry rates tend to be highest, since major e-commerce platforms invest heavily in anti-bot detection and will aggressively block or challenge requests.

Consider the alternative: at $499/month on an unmetered plan, you could monitor 10,000 SKUs every 15 minutes instead of every hour, quadrupling your data freshness while still paying less than the cheapest per-GB option. That's the kind of operational freedom flat-rate pricing unlocks.

When Per-GB Actually Makes Sense

Fairness matters. Per-GB isn't always the wrong choice.

If you run less than 10GB per month, per-GB pricing can genuinely cost less. That covers use cases like light API testing where you're making a few thousand requests, occasional one-off scraping jobs that run for a day and stop, low-volume research where you're pulling a few hundred pages per week, or early-stage projects where you're still validating whether proxies are the right approach for your workflow.

The break-even point sits around 10-20GB per month. Below that range, per-GB keeps your costs low because you're only paying for what you use. Above it, unmetered starts saving you money. And the further above you go, the more dramatic the savings become. At 100GB+, the difference is often thousands of dollars per month.

The question to ask yourself: is your workload growing? If you're at 15GB today but expect to hit 50GB next quarter as you scale your operation, locking in an unmetered plan now avoids the sticker shock later. If you're genuinely running small, occasional jobs with no growth trajectory, per-GB keeps things simple and predictable.

The Hidden Costs Nobody Mentions

The per-GB vs unmetered comparison isn't just about the invoice. There are costs that never show up on a bill but still hit your operation.

Bandwidth anxiety. When every megabyte has a price tag, you make different decisions. You skip retries to save data. You reduce scraping frequency from hourly to every six hours. You avoid loading full pages and try to parse partial responses instead. You throttle your own operation to stay within budget, even when the proxy could handle more. The proxy works fine. You're the bottleneck.

Unpredictable bills. Per-GB pricing means your costs scale with usage, which sounds rational until you realize usage is hard to predict. A target site changes its structure and your scraper suddenly downloads 3x more data per page. A social platform pushes heavier feeds. A competitor launches a flash sale and you need to increase monitoring frequency. Each of these is a surprise invoice waiting to happen.

Optimization overhead. Engineering time spent minimizing bandwidth is time not spent building features or expanding coverage. When your team is debugging response sizes, compressing requests, writing custom retry logic to reduce data waste, and building dashboards to track bandwidth consumption per job, they're optimizing for the wrong metric. The goal should be getting the data you need, not getting it in fewer bytes.

Failed request waste. This one stings the most. With per-GB pricing, a request that returns a CAPTCHA page, an error, or a block still counts as bandwidth consumed. You pay full price for data that gave you absolutely nothing useful. On a bad day with high block rates, 40-50% of your bandwidth bill might be pure waste. You're literally paying for the privilege of being blocked.

What to Look for in an Unmetered Provider

Not all unmetered plans are created equal. "Unlimited bandwidth" is a marketing term that plenty of providers use while hiding significant limitations in the fine print. Here's what separates a real unmetered plan from a dressed-up one.

Published speed formula. Your throughput on an unmetered plan depends on thread count and per-thread speed. If the provider won't publish how speed is calculated, you can't predict your actual performance. Look for a clear formula. At ScaleProxy, it's threads x 2 = max Mbps. Starter (100 threads) gets up to 200 Mbps. Growth (500 threads) gets up to 1 Gbps. Scale (1,000 threads) gets up to 2 Gbps.

No fair use policy. Some providers advertise "unlimited" bandwidth but bury a fair use clause in their terms. Once you exceed an undefined threshold, they throttle your connection or suspend your account. If the word "fair use" appears anywhere in the terms of service, the plan isn't truly unmetered.

Uptime SLA with teeth. An uptime guarantee only matters if it has automatic enforcement. "99.9% uptime" means nothing if claiming a credit requires opening a ticket, proving the downtime, and waiting weeks for a review. Look for providers that issue credits automatically when the SLA is missed. No ticket required.

Real migration support. Switching providers is a hassle. Configuration changes, IP pool differences, session handling quirks. A provider that offers hands-on migration support, someone who will walk you through the transition and help troubleshoot configuration issues, makes the switch significantly less painful. This matters more than most people realize until they're mid-migration at 2 AM wondering why their sessions keep dropping.

The residential proxy market has dozens of providers, but the pricing model you choose matters more than most features on a comparison sheet. Per-GB works at small scale. Unmetered works at real scale. The math doesn't lie.