Almost every pricing page for residential proxies leads with one number: the size of the pool. Millions of addresses, sometimes tens of millions, sometimes more. It is the most prominent figure in the market and the hardest to check.
If you are choosing a residential proxy plan, the pool size on the box tells you less than it appears to. This guide explains what that number usually counts, what you get instead, and how Atlas measures the part you can verify.
What an advertised pool size counts
There is no standard definition of an IP pool size, and no auditor checks the figure. In practice it is usually a cumulative count: every distinct address the provider has seen on its network over a month or longer, across every country it serves.
That figure can include:
- Addresses that were online once and have not been seen since
- Addresses in countries you will never request
- The same household counted several times as its address changed
- Addresses the provider could reach through a partner, not its own sourcing
None of this is necessarily dishonest. A cumulative count is a reasonable way to describe how much inventory a network touches. It is just not the same as how many addresses will answer your requests. The methodology puts it bluntly: advertised pool size describes inventory, and it is unverifiable by construction.
What you actually get is a sample
At any given moment, only part of a residential pool is online. Devices go to sleep, households switch off their routers, and addresses get reassigned. Of what is online, only the slice in the country you asked for is available to you. Of that slice, you see only as many addresses as your request volume and the gateway’s rotation allow.
So the useful question is not “how big is the pool” but “how many distinct addresses answered my requests, in my country, in a given window”. That count can be checked, and it can be compared across providers when they are tested the same way.
Why observed counts look smaller than the box
Take an illustrative example. A provider advertises ten million addresses. Say a third of them are online at a typical moment, and a tenth of those are in the country you need. That leaves a few hundred thousand addresses you could plausibly reach.
Now send rotating requests. The first thousand will mostly come back from new addresses. By the hundred-thousandth, a growing share will be addresses you have already seen. Plot distinct addresses against requests sent and the line rises steeply, then bends and flattens.
Where that curve flattens is the real size of the pool for your country, as the gateway serves it. It is almost always far below the headline, and that is expected. A curve that flattens early, after relatively few requests, is the warning sign: it means the gateway keeps handing you the same small set.
How Atlas counts unique IPs
Atlas counts distinct exit addresses actually observed answering its requests. A few rules make that number comparable:
- The target’s view. The address recorded is the one the target saw, never a figure from a provider dashboard.
- Same schedule for everyone. Every network gets the same requests, spread evenly across the window and across targets.
- The product as sold. Session and rotation settings stay at each provider’s defaults.
- Fixed windows. Pool size is the distinct exits seen in the last 24 hours, 7 days or 30 days, as set out in the window rules for every figure.
The result is a floor, not a census. It shows what a network delivered to a steady, realistic workload. Because the method is identical for every network, the unique IP counts on the networks page can be compared directly, which no two advertised pool sizes can.
Size is not quality
A big observed count is good, but it is not the whole story. Three other figures change what a count is worth:
- Network spread. A pool concentrated on a handful of autonomous systems fails together when one of them is blocked. The ASN diversity guide covers how to measure it.
- Hosting addresses. Addresses from data centres inside a pool sold as residential inflate the count while getting flagged by targets. See how to tell real from fake residential proxies.
- Repeat addresses. A pool whose addresses come back day after day behaves more like a small fixed set than a large rotating one.
A smaller pool of clean, well-spread addresses in the right country can outperform a much larger one on every target that matters to you.
Pool sizes do not add up
Buyers often run two providers and assume the pools combine. If one network claims two million addresses and another claims one million, three million feels like the total. When the second resells part of the first, the real figure is much closer to two million.
The only way to know is to compare the actual exit addresses, tested at the same time. The guide to why two proxy providers are often the same provider explains how that comparison works and how to read its direction.
Doing it yourself
You can estimate a pool’s real size for your own use in an afternoon:
- Pick one country and one endpoint that echoes back the address it saw.
- Send a fixed number of rotating requests at a steady rate, at a time of day you would really run jobs.
- Every thousand requests, record the number of distinct addresses seen so far.
- Plot the curve. Note where it starts to flatten.
- Repeat on a different day. A healthy pool shows many new addresses on day two.
- To compare providers, run them side by side in the same window with the same request count.
Questions to ask a provider
If a provider quotes a pool size, these questions turn it into something you can use:
- Over what period are addresses counted?
- How many are online in a typical hour, in the country I need?
- Does the figure include addresses reached through partners?
- Are hosting or data centre addresses excluded?
- Will you let me test before I commit, so I can count for myself?
A provider confident in its pool will answer most of these. The last one matters most, because a count you ran yourself beats any figure on a pricing page.