We test residential proxies from 18 providers every minute of every day. These guides explain what the numbers mean, how to read a provider's pricing page, and how to keep a scraping job running without burning bandwidth on blocked requests.
Twelve checks you can run during a trial, each with a test procedure and the answer that should end the conversation. Copy the scoring table and fill it in as you go.
Read the guideRouting traffic through a rented home IP address breaks no law by itself. The exposure comes from the request you send, the data you keep, and how the household on the other end was recruited.
Read the guideGateway usernames carry your country, city and ASN filters, and every step down that ladder throws away part of your pool. Here is what each level is worth and how to check what you actually received.
Read the guideChallenges appear when a confidence score falls below a site-specific threshold. The exit IP is one input among several, which explains both why a better pool helps and why it never removes the problem entirely.
Read the guideA 403, a 429, a CAPTCHA and a silent empty 200 are four different problems with four different fixes. Identify which one you have before you change a single header.
Read the guidePicking a pool is the short part of the job. This is the wiring: gateway credentials, session identifiers, retries that do not multiply your bill, and the instrumentation that shows which exits are failing.
Read the guideEvery image your crawler downloads through a residential exit has a price on it. These are the configuration changes that remove two thirds of the bill while visiting exactly the same pages.
Read the guideHow proxies work, what the different types are, and which one fits your job.
Rotation is a property of your proxy username, not of your code. Here is what each mode does to a request, the failure signature of each, and the workload table that settles the argument.
Read the guideOne protocol reads your request and can rewrite it. The other moves bytes and asks no questions. The choice changes your DNS path, your UDP options and your handshake cost.
Read the guideCarrier NAT puts thousands of paying subscribers behind the address you are borrowing, which makes banning it expensive for the target. You pay handsomely for that shield.
Read the guideA machine in a rack, announcing address space leased from a consumer internet provider. That combination is what you are buying when a vendor sells you static residential IPs.
Read the guideOne lookup separates these two products: the organisation that owns the exit IP. Everything else, including the tenfold price gap, is downstream of that.
Read the guideResidential proxies borrow IP addresses that ISPs handed out to real households. That single fact explains their price, their latency, and why anti-bot systems treat them differently from server IPs.
Read the guidePricing models, pool size claims, and how to evaluate a provider before you pay.
Twelve checks you can run during a trial, each with a test procedure and the answer that should end the conversation. Copy the scoring table and fill it in as you go.
Read the guideEvery image your crawler downloads through a residential exit has a price on it. These are the configuration changes that remove two thirds of the bill while visiting exactly the same pages.
Read the guideSomebody is paying for the bandwidth on those open ports. This is what they get in return, described mechanically, plus the short list of jobs where the trade is still acceptable.
Read the guideA hundred million addresses is a cumulative census of an entire planet over a long window. Your crawl is limited by how many distinct exits answer in one country inside one rotation window.
Read the guideTwo providers can advertise the same dollar figure per gigabyte and still send you very different invoices. The gap hides in the tier table, the expiry clause and your own success rate.
Read the guideWhat success rate, latency, fraud score and clean rate actually measure.
Run every provider at the same time, against the same targets, from the same machine, and log enough that you can compute cost per successful request when the trial ends.
Read the guideA residential proxy adds two uncontrolled network legs to every request, so speed has to be measured statistically. Here is the curl harness, the percentile maths and the sampling rules.
Read the guideFraud scores are model output, not measurement. Understanding which signals feed them tells you when a low score will get you through a checkout and when it buys you nothing at all.
Read the guideA provider quoting 99.9% and a scraper seeing 78% can both be telling the truth. The difference is the predicate, the denominator and the target, and all three are yours to fix.
Read the guidePractical tactics for keeping large scraping jobs unblocked and cheap.
Gateway usernames carry your country, city and ASN filters, and every step down that ladder throws away part of your pool. Here is what each level is worth and how to check what you actually received.
Read the guideChallenges appear when a confidence score falls below a site-specific threshold. The exit IP is one input among several, which explains both why a better pool helps and why it never removes the problem entirely.
Read the guideA 403, a 429, a CAPTCHA and a silent empty 200 are four different problems with four different fixes. Identify which one you have before you change a single header.
Read the guidePicking a pool is the short part of the job. This is the wiring: gateway credentials, session identifiers, retries that do not multiply your bill, and the instrumentation that shows which exits are failing.
Read the guideWhere residential IPs come from and what you are allowed to do with them.
Routing traffic through a rented home IP address breaks no law by itself. The exposure comes from the request you send, the data you keep, and how the household on the other end was recruited.
Read the guideEvery guide here is written against the same dataset that powers our public rankings. Success rate, latency, fraud score and price per 100GB are re-measured continuously.