"Rotating proxies" means quite different things from one provider to the next. Two offers can carry the same label and behave nothing alike once your team's tools are pointed at them. This guide covers what happens behind the gateway, which settings change your results, and what to measure before anyone signs off on spend.
What rotation looks like on the network
Your tools connect to one gateway, often called a backconnect endpoint (for example gate.example.com:7000). Behind it, the provider holds a pool of thousands or millions of outgoing IPs, and each request leaves through a different one.
The pool behind the gateway can be made up of:
- Datacenter IPs: thousands of addresses owned by the provider.
- Static residential (ISP) IPs: thousands of addresses registered to consumer internet providers. With us, these come with Residential Premium and stay fixed for the life of the order instead of rotating.
- Residential IPs: millions of household addresses from a consent-based network.
- Mobile IPs: thousands of addresses from mobile carriers.
Per-request or sticky: the session settings
Two modes cover almost every business job:
- Per request: a different IP every time. Built for volume, with no continuity between requests, which is right for price and rank checks.
- Sticky session: you pass a session ID and the gateway keeps every request with that ID on one IP for a set window, often 1-30 minutes. Right for signed-in work and multi-step checks.
The session ID usually travels in the username as session-{id}, like this:
user-abc123-session-xyz789:password@gate.example.com:7000Keep the ID and you keep the IP; change it and you get a new one. Many gateways read the location from the username as well:
user-abc123-country-us-city-newyork-session-xyz789:password@gate.example.com:7000Pool size: ask for the number that applies to you
A headline such as "100 million IPs!" says little about your project. Ask these questions instead:
- How many IPs are online in your target country during your working hours? If you need US residential and only 5,000 are live in that window, that is your real pool.
- How clean are they? IPs with a recent history of abuse get blocked sooner, and large shared pools carry more of that history.
- How many networks do they come from? If all 5,000 sit on Comcast and a site blocks Comcast for a day, the whole job stops.
Budgeting for data
Across the market, rotating residential is usually billed per GB, while some providers sell datacenter by IP or by port. We bill all four of our types per GB, so the real question for any team is how much data its jobs use. Even a moderate monitoring job can use 10-50 GB a month. Estimate yours with the proxy calculator before the first order.
What to measure before you commit
- Success rate on your own target: send 1,000 requests and count the 200, 403 and 429 responses and the CAPTCHAs.
- Unique IPs in practice: run 100 IP lookups and count how many different addresses come back.
- Latency: residential answers more slowly than datacenter or static IPs, so make sure the difference suits your schedule.
- Location accuracy: when you ask for Germany, confirm you get Germany with our IP lookup tool.
Mistakes that waste rotation
- Rotating per request while signed in: the session breaks.
- Sharing a single session ID across every job: you lose the variety you are paying for.
- Ignoring 429 responses: one overloaded IP can hold up a job even with rotation. Back off, start a new session and retry.
- Carrying cookies across IPs: the site sees one signed-in user hopping between addresses and may lock the account to protect it.
How rotation works on our plans
All our proxy plans offer per-request rotation and sticky sessions with 1-30 minute windows, country targeting, and city targeting where it is available. Over SOCKS5, DNS lookups also go through the proxy, so they don't give away your own network. Neither mode costs extra. To pick between the two task by task, see sticky vs rotating sessions.
