Sticky vs rotating IP: how mobile proxy rotation actually works

Here is the short version before the detail. Use a sticky IP when the job has to be remembered as one person, a login, a checkout, an account session. Use a rotating IP when each request stands on its own, price checks, search results, wide scrapes. Pick the wrong one and the proxy works perfectly while quietly ruining the job.

I run the hardware these IPs come from, real phones on real SIMs, so I set up both of these every day. The good news is this is not a tier you pay more for. It is a port you connect to.

It is a port, not a product

The first thing to unlearn is that sticky and rotating are two products. They are two endpoints on the same line. One modem on my farm gives you a sticky port and a rotating port. Same phone, same SIM, same credentials, you just point at the one that matches the job.

That means the real decision happens before you connect, not after you buy. If you are fuzzy on how the host, port, and credentials fit together, what is a mobile proxy covers the anatomy. Here we are only deciding which port.

So the question is never “which should I buy.” It is “which does this specific task need.” And that comes down to one thing: does the work need to be remembered as a single person, or is each request a stranger.

When sticky wins

Sticky holds one IP for a window, long enough to finish something that has to happen on a single identity. Anything with a session needs this.

A login is the clearest case. You send a username, a password, sometimes a 2FA code, and the site watches all of it come from one address. A checkout is the same, you browse, add to cart, and pay across several requests that need to look like one shopper. Account warm-up, posting, filling a multi-step form, all of it wants sticky.

The rule is simple. If a human would do it in one sitting on one phone, use sticky. The whole point of running multiple accounts without bans is that each account looks like one consistent person, and a stable IP is half of that picture.

When rotating wins

Rotating gives you a fresh IP on the carrier schedule, which is exactly what you want when each request is independent and you do not want a thousand of them coming from one address.

Price checks across a catalog, search results, gathering listings, any wide read where request number 900 has nothing to do with request number one. Spreading that load across many carrier IPs is what keeps you under per-IP rate limits and looking like normal traffic instead of one machine hammering a site. Most of the jobs in these web scraping case studies are rotating jobs for exactly that reason.

The mental model: rotating is for strangers. Every request can be a different person on a different phone, because none of them need to remember the others.

The two failure modes that catch everyone

Almost every “my proxy is broken” message is really one of these two.

Rotating mid-login. You connect to the rotating port, start a login, and the IP changes between the password and the 2FA step. The site sees one person turn into two and flags the account. The proxy did its job, it rotated, you just asked it to rotate at the worst possible moment. Logins are sticky work.

Sticky for a giant crawl. You connect to the sticky port and fire ten thousand requests through one held IP. That single address carries all the load, trips the rate limit, and starts getting 429s. The proxy held the IP exactly like you asked, you just asked one IP to do a job meant for many. Big crawls are rotating work.

Both look like a faulty proxy. Neither is. You picked the wrong port for the task.

Mixing both in one workflow

Real setups often need both, and that is fine because both live on the same account. A common shape: rotating for the anonymous reads, sticky for the logged-in actions.

Say you are researching listings and then acting on a few from a logged-in account. Pull the listings on the rotating port so the wide read spreads across IPs. Then switch to the sticky port for the logged-in part so that session stays on one address. Two ports, one account, each task on the right one.

You do not need two subscriptions for this. You need to point each part of your code at the right endpoint.

How rotation actually times itself

One thing that trips people: a rotating IP is not random. It moves within the carrier’s own range on a schedule, the way every real phone gets a new address as the network manages its pool. Some setups also let you force a fresh IP on demand. Either way, you are talking to a stable gateway, and the IP behind it changes. You never hardcode the raw rotating IP, because the whole design is that it moves.

Sticky is the same gateway holding the address still for your window instead of cycling it. Knowing your sticky window matters, it is minutes to about an hour depending on the line, not forever. Finish the session inside it.

The one rule to remember

Strip away everything else and you get one question.

Does this task need to be remembered as one person? Use sticky. Is each request a stranger that owes nothing to the others? Use rotating. That single distinction picks the right port for almost everything you will ever do.

If you want to feel the difference instead of reading about it, there is a free trial of Singapore Mobile Proxy, real SIMs on Singtel, M1, and StarHub from my own farm. You get both a sticky and a rotating endpoint, so you can run a login on one and a wide pull on the other and watch exactly how each behaves in the first ten minutes.