How residential proxies actually work
When you buy residential proxy bandwidth, your scraper's request does not reach the target site from your server. It exits through a real household's internet connection, somewhere else in the world, and arrives carrying that home's IP address. Everything else about this product follows from that one mechanic.
The path a request takes
Your code sends the request to the provider's gateway rather than to the target. The gateway picks an available exit device from its pool (a phone, a laptop, a smart TV) and relays your request through it. The target site sees a residential connection from a consumer ISP, because that is literally what it is. The response comes back the same way.
This is why residential bandwidth costs roughly ten to fifty times what datacenter bandwidth costs. You are not paying for servers. You are paying for access to a network of other people's connections, and for the software that keeps enough of them online at once.
Who the household is
Someone installed something. That is the entire supply side of this market, and providers differ almost entirely in what that something was and what the person got for it.
The three models below are not interchangeable, and the difference between them is the difference between a paid transaction and a checkbox in the terms of a free game.
Why the pool number is soft
Every provider advertises a pool size: 85 million addresses, 150 million. Treat these as marketing rather than measurement. They generally count every address seen over some unstated window, not addresses available right now, and a device that was online for ten minutes last March counts the same as one that is online today.
We render these as the vendor's own claim, labelled as theirs, because there is no way for us or anyone else outside the provider to verify them. A number we cannot check is shown as a number we cannot check.
What actually determines whether it works
Success rate against your specific targets, which no comparison table can tell you, including this one. Two providers with identical pool claims and identical prices can differ enormously on a given site, because what matters is whether that site's anti-bot vendor has already seen and flagged those particular addresses.
This is why a trial is worth more than a specification sheet, and why we have not declared a winner.
Bundled SDK
A software development kit is embedded in a free app (a VPN, a game, a utility) and the app's terms mention that spare bandwidth may be shared.
The household: Gets the app free. Usually is not paid.
The catch: Consent lives in a terms-of-service screen almost nobody reads, on an app installed for an unrelated reason. This is the model the FBI's Popa botnet reporting describes at scale.
Paid app
A standalone app pays people to share their connection. Installing it is the entire point of the app.
The household: Is paid, typically a few dollars a month.
The catch: Consent is explicit and the transaction is understood. The person still usually cannot see what traffic exits through their line, or refuse a particular use of it.
ISP agreement
Addresses are obtained through a commercial arrangement with an internet service provider rather than from end-user devices.
The household: Is generally not involved and not paid.
The catch: Cleaner on consent, and it is the model NetNut claimed. The FBI seized it anyway. The sourcing label is a claim, not an audit.
Datacenter
Addresses belong to servers in a datacenter. No consumer connection is involved at any point.
The household: Does not exist. There is no household in this model.
The catch: No consent question at all, which is the honest reason to prefer it where the target site tolerates it.
Questions
Is my traffic visible to the person whose connection it exits through?
No. The traffic is encrypted end to end between you and the target site, and the exit device relays it without being able to read it. The household generally cannot see what is passing through, which is also why they cannot object to a particular use of it.
Why are residential proxies so much more expensive than datacenter?
Because you are paying for access to consumer connections rather than for servers. Datacenter addresses are cheap to create and easy for target sites to identify in bulk; residential addresses are neither.
Are advertised pool sizes real?
They are unverifiable from outside the provider, and they generally count addresses seen over an unstated window rather than addresses available now. We show them as the vendor's own claim, labelled as theirs.
The rest of this section
- Are residential proxies legal?
- Bundled SDK or paid app: what the household actually agreed to
- Where residential proxy IP addresses actually come from
- What regulators and prosecutors have actually done
- The FBI seized NetNut. What it means if you were a customer, what actually happened
These pages carry no affiliate links, ever, including to providers they name. How this site pays for itself.