Bright Data vs Oxylabs vs Decodo: which residential proxy actually fits

Bright Data, Oxylabs and Decodo are the three residential proxy providers most scraping shortlists come down to, and the instinct is to rank them by price per GB. That is the wrong sort first. List rates move, differ by region, and collapse once you commit monthly volume, which is exactly why five comparison pages will quote you five different numbers for the same provider and all be honest. The real differences are billing model, whether you need a managed unblocker, and how fast you can start.

The three, in one line each

• Bright Data: the largest and most feature-heavy network. Per-GB residential proxies plus a managed unblocker (Web Unlocker) and a Scraping Browser, at a premium price, with KYC-first onboarding.
• Oxylabs: the other enterprise incumbent. A similar premium tier, per-GB residential, its own scraper APIs and unblocker, and the same compliance-heavy onboarding.
• Decodo: rebranded from Smartproxy in April 2025. Self-serve, cheaper per GB, aimed at small-to-mid jobs, and expanding from a proxy service into a broader scraping platform.

What the per-GB price actually is, and why comparisons disagree

Line up several comparison sites and you get several per-GB numbers for the same provider, because pay-as-you-go list rates move, vary by region, and fall once you commit volume. As a rough shape most sources agree on: the premium pair list around 8 to 12 dollars per GB pay-as-you-go and fall toward the low single digits only at high commitment; Decodo lists meaningfully below that and reaches roughly 2 to 3 dollars per GB at terabyte scale; and a budget tier, IPRoyal, Webshare and similar, starts under about 3 dollars per GB. (List rates, checked July 2026, they move.) Treat any single headline figure as a starting point, not a quote, and price your real volume.

The line most comparisons skip: onboarding and KYC

Bright Data and Oxylabs both take a KYC-first approach to residential access, and higher-volume residential plans, commonly reported around the 500-dollars-a-month mark, typically trigger a KYC call and a compliance review before your traffic is switched on. That is a feature, not a flaw: it is part of what keeps the pool clean and the vendor on the right side of data-protection law. But if you need to start scraping this afternoon, it is a real difference from the more self-serve Decodo, which also runs KYC but gates less of its entry tier behind it. Factor onboarding time in, not just price per GB.

Are residential proxies legal, and where do the IPs come from

This is the axis the price comparison skips, and it is the one that carries the real risk. A residential proxy routes your request through a genuine home or mobile connection, and those IP addresses belong to the internet service providers and the people using them, not to the proxy vendor. So the whole model rests on how the vendor got permission to lend them out. The lawful and ethical way is informed consent: an app or SDK where the device owner opts in, usually in exchange for an ad-free tier, a premium feature or a small payment, and can leave at any time. The premium pair document exactly this. Bright Data publishes a sourcing trust center and commits that a peer can opt out and have the app and SDK removed from the device; Oxylabs publishes its own ethical-sourcing and acceptable-use standards. That paperwork is what the KYC call above exists to protect.

The other end of the market is where the trouble sits. Some pools are built from devices infected by malware or bundled into a free VPN whose users never understood what they agreed to. The clearest case is the 911 S5 network the US Department of Justice dismantled in May 2024: roughly 19 million residential IPs across more than 190 countries, silently installed through free VPN apps without consent and used for large-scale fraud, with the administrator arrested and the operators sanctioned. For a buyer the exposure lands in two places. First the sourcing: routing your traffic through IPs obtained without consent can pull you into someone else liability, which is why a free residential proxy list, almost never ethically sourced, is never worth the saving. Second the data: scraping public, non-personal data through a consented pool is standard and lawful in most jurisdictions, but collecting personal data, or anything behind a login, brings GDPR and computer-misuse law into play whatever proxy you used. So make a written sourcing disclosure part of the shortlist, not an afterthought. The KYC-gated vendors gate their pool for the same reason you should care which pool you are buying.

The cost the per-GB rate hides: does the bandwidth expire

There is a second reason cheapest-per-GB can mislead, and it sits in how you buy the gigabytes, not their sticker rate. The premium pair sell two ways that behave very differently. A monthly committed plan buys an allowance at a lower per-GB rate, but it is a minimum, not a wallet: Bright Data states its monthly plans are minimum commitments and unused gigabytes do not roll into the next month, so commit to a 500-dollar plan and use a third of it and you still paid the full 500, which puts your effective cost at three times the sticker. Pay-as-you-go removes the commitment but charges the top rate, around 8 to 12 dollars per GB on the premium tier. And the budget tier is where this flips outright: IPRoyal and DataImpulse advertise residential bandwidth that does not expire and rolls until you consume it, while Oxylabs does not offer that non-expiring rollover, so prepaid data you never use is money spent. Decodo sits between as a self-serve monthly subscription, so check its current plan terms rather than assuming either way.

The practical read is that the number deciding the bill is cost per gigabyte you actually use, not the headline per-GB rate. A committed plan is cheaper per GB only if you reliably burn most of it every month; for spiky or exploratory work a higher pay-as-you-go rate, or a budget provider whose bandwidth does not expire, can be cheaper in practice than a low-rate plan you never fully consume. Size your real monthly volume first, then choose the buying model, because the plan shape can move your effective rate more than the choice of provider does. (Plan terms, checked July 2026, they move.)

Billing model matters more than the sticker

All three sell residential proxies by the gigabyte, so on raw residential the cheapest-per-GB provider wins a job where you already know the target lets a plain proxy through. The moment the target fights back with fingerprinting and CAPTCHAs, the useful product is not a raw proxy but a managed unblocker that bills per successful request and folds retries in. Bright Data (Web Unlocker, Scraping Browser) and Oxylabs both sell that layer; it costs more per record but carries the block-handling code you would otherwise write and keep alive as the target changes. On a hard target the unblocker, not the per-GB rate, is the number that decides the bill.

The tax the meter hides on hard targets: failed requests still bill

There is one more reason the cheapest-per-GB provider can lose on a defended target, and it is the least visible: a raw residential proxy meters bytes, not successes. Providers bill the sum of everything that crosses the proxy, request headers, any POST body, response headers and response body, so the HTTP status code is irrelevant to the meter. A block page, a CAPTCHA challenge, a 403 or a compliance error all return real bytes, and those bytes count. Bright Data documents its residential bandwidth exactly this way, as the total data transmitted to and from the target, and a blocked request is still data transmitted.

That turns every retry into another full charge. On a hard target where one request in three comes back blocked and your scraper retries it twice before it gets through, you have paid for three responses to bank one, so your real cost per useful record is roughly triple the page weight, before you switch provider or trim a single image. This is the mirror image of the managed unblocker above: it lists a higher unit price but bills per successful request and absorbs the retries, which is exactly why it can come out cheaper per delivered record on a target that fights back. So on anything defended, the honest figure is gigabytes, or euros, per successful record, not the sticker per GB and not the page weight of a clean fetch. Meter your block rate first: if a large share of requests are failing, the fix is a better tier or an unblocker, not a lower per-GB rate on traffic you are already burning on block pages.

The biggest lever on a per-GB bill is what you download

Because the meter is gigabytes, the largest saving on the table is usually not the cheapest-per-GB provider, it is downloading fewer gigabytes. Most of a page is not the data you came for. The 2025 median web page is about 2.9 MB on desktop, and of that images are roughly 1.06 MB and JavaScript roughly 0.7 MB, while the raw HTML is only about 22 KB (HTTP Archive Web Almanac 2025). If the value you are scraping lives in that HTML, rendering the full page through a residential proxy means paying premium rates for something close to a hundred times the bytes you actually need.

There are two levers, and both cut the bill before you ever compare providers. First, fetch the HTML directly rather than driving a headless browser, when the data is in the server-rendered markup or reachable through a JSON or XHR endpoint the page already calls; you skip the images, fonts and scripts entirely. Second, when the site is a client-side app you genuinely have to render, block images, media, fonts and analytics at the request level so the browser never pulls them through the proxy. Either one can strip most of the page weight, and on a per-GB bill that is a bigger lever than switching from one provider to another.

The caveat is that some hard targets fingerprint on whether a real browser loads its assets, so blocking resources too aggressively can itself get you flagged; that trade-off is part of what a managed unblocker is handling for you. So measure gigabytes per successful record, not just the sticker per-GB rate, and price the real page weight you download rather than the full page. The calculator estimates your monthly GB from exactly that page weight, so trimming what you fetch lowers the number it quotes.

Where each one wins

• Bright Data: the hardest targets and the widest toolset. Reach for it when you need the managed unblocker or the largest network and can absorb the premium price and the KYC onboarding.
• Oxylabs: the enterprise alternative at a similar tier. A fit when you want that scale and support from a different vendor, or its scraper APIs suit your stack better.
• Decodo: the value pick for small-to-mid jobs. Cheaper per GB, self-serve and fast to start, which is often all a lightly defended target needs.
• The budget tier (IPRoyal, Webshare and similar) and self-hosted datacenter proxies are cheaper still, and genuinely enough when the target barely defends itself. Do not pay residential rates for a site that never asked for them.

Pick the tier before the brand

The recurring mistake is shortlisting on brand and headline price before knowing which tier the target forces. A weakly defended site runs fine on cheap datacenter or budget residential IPs; a hard one needs a managed unblocker, and there the brand is a rounding error next to that choice. Put your real page count, page weight and target difficulty into the calculator: it estimates your monthly GB, picks the tier that actually gets through, and prices it, so you shortlist providers for the tier you actually need instead of the one with the loudest homepage.