More unique IPs than any network on the planet, in the world's largest residential pool. That's Shifter.More unique IPs than any network, in the world's largest pool. That's Shifter.

See the numbers

Why two proxy providers are often the same provider

How to tell whether a proxy network owns its inventory or resells someone else's, why that matters for redundancy, capacity and pricing, and how to measure address-level overlap without fooling yourself.

Elena Petrova Data Science · Updated November 13, 2026 · 11 min read
Two pools of exit IPs. Amber addresses appear in both.
Industry
On this page
  1. How residential supply actually works
  2. The three decisions it breaks
  3. Direction is the whole finding
  4. The sampling trap
  5. Signals beyond shared addresses
  6. What overlap does not prove
  7. What a second provider really buys you
  8. How to use it when buying
  9. Doing it yourself
  10. How Atlas measures it
  11. Frequently asked questions

The residential proxy market has a structural feature that is rarely discussed in public: many of the “networks” you can buy from do not own most of the addresses they sell. They buy wholesale from a smaller number of operators and resell a slice, sometimes with real added value in routing, targeting and tooling, sometimes with nothing added but a margin.

This is not a scandal. Resale is normal in every infrastructure market, from mobile networks to cloud hosting. The problem is that it is invisible at the point of purchase, and it quietly breaks three decisions buyers make all the time: which provider to use as a fallback, how much capacity they really have, and what a fair price is.

The good news is that it can be measured. This guide explains how the supply chain works, what overlap does to those decisions, and how to measure it without the sampling mistakes that make most overlap figures meaningless.

How residential supply actually works

A residential exit is a real household connection that has agreed, usually through an app or SDK, to route traffic for a network. Building that supply is slow and expensive: it takes app partnerships, consent flows, payouts, fraud control and years of churn management. Only a handful of operators do it at scale.

Around them sits a layered market:

  • Operators source the devices and run the infrastructure that routes traffic through them.
  • Wholesalers buy capacity from one or more operators and sell it on in bulk, sometimes blending several sources behind one gateway.
  • Resellers buy from an operator or a wholesaler and sell under their own brand, often with their own dashboard, plans and support, while the traffic leaves through someone else’s exits.

A single brand can sit in more than one layer: own supply in some countries, bought-in supply in others, and a wholesale feed to fill gaps during demand peaks. None of this shows on a pricing page.

The three decisions it breaks

Redundancy. The standard advice when a provider fails you is to keep a second one. If both draw from the same pool, you have bought a second invoice rather than a second source. When a target starts refusing a block of addresses, it refuses them for both providers at once, which is exactly the moment the fallback was meant to help. If 70% of your fallback’s exits are shared with your primary, 70% of your fallback traffic lands on addresses the target has just learned to distrust. You usually find this out during an incident.

Capacity. Providers advertise pool sizes, and buyers add them up. Suppose a large network shows you 1,000,000 distinct exits and a smaller one shows you 70,000, of which 50,000 also appear in the large one. Running both gives you 1,020,000 distinct exits, not 1,070,000. The second provider added 2% more addresses, not 7%, and you pay for its whole plan.

Price. A reseller’s cost is its upstream’s price plus its own margin. That can be fine if the reseller adds something you value. It is poor value if you could buy the same exits upstream for less, and many buyers do not know an upstream exists.

Direction is the whole finding

A single overlap percentage hides the most useful part of the measurement. Shared addresses have to be read from both sides.

Take the example above: 50,000 addresses in common between a network with 70,000 observed exits and one with 1,000,000.

Small networkLarge network
Distinct exits observed70,0001,000,000
Shared with the other50,00050,000
Share of its own pool that is shared71%5%
Exclusive share29%95%

That shape, with most of the small pool inside the large one and only a sliver of the large one inside the small, is the signature of resale. The relationship runs one way: the small network is downstream.

Compare two independent operators sourcing from consumer ISPs in the same countries. Say one shows 400,000 exits, the other 300,000, and they share 6,000. That is 1.5% of one and 2% of the other: a small overlap in both directions, roughly what you would expect from two networks that happen to reach some of the same households. Symmetric and small means two networks doing the same thing separately. Asymmetric and large means one is buying from the other.

The sampling trap

Here is the mistake behind most overlap figures you will see: a small sample makes even identical pools look unrelated.

Imagine two brands that sell exactly the same 2,000,000-address pool. You sample 20,000 exits from each. On average, only about 200 will match, about 1% of each sample. Read naively, that says the two networks barely overlap. In fact they are the same network.

This happens because a rotating pool hands out a different slice every time you look. Raw overlap is always a floor: two pools can share far more than you saw, never less. The question is how to tell “different pools” from “same pool, small sample”. There are two ways.

Compare with each pool’s overlap with itself. Take two samples from the same provider, in the same window and the same way, and measure how much they overlap with each other. That is the baseline: the most overlap you could expect to see between two samples of one pool. Then compare the cross-provider overlap with it.

relative overlap = overlap between A and B ÷ overlap of A with itself

A relative overlap near 1 means B’s exits are, as far as your samples can tell, drawn from A’s pool. Near 0 means independent. In between means partial sharing, such as a network that buys some of its supply.

Estimate the real pool size from repeat sightings. The same self-overlap gives you a rough estimate of how many addresses a provider can really reach, using the capture-recapture method ecologists use to count fish in a lake:

estimated pool size ≈ (sample 1 size × sample 2 size) ÷ addresses seen in both

If two samples of 40,000 exits from one provider share 8,000, the provider is reaching roughly 200,000 distinct addresses on your targets in that window, whatever its marketing says. The estimate assumes every address is equally likely to be handed out, which is never quite true, so treat it as an order of magnitude rather than a census. It is still far closer to reality than an advertised pool size.

Signals beyond shared addresses

Address overlap is the proof. A few other signals point the same way and are often easier to spot first:

  • Shared incidents. If two networks start failing on the same target in the same hour, and recover together, they are probably routing through the same exits. Correlated incidents over weeks are hard to explain any other way.
  • The same location errors. Every network maps addresses to countries and cities with some error. When two networks get the same addresses wrong in the same way, they are likely reading the same upstream location data, or sending the same traffic.
  • Identical gateway behaviour. The same error pages, the same session parameter syntax, the same quirks in sticky sessions or the same response headers from the proxy layer suggest the same infrastructure underneath.
  • Matching ASN and country mix. Two pools with nearly identical shares by ASN and country are suggestive. On its own it proves nothing (see below), but next to an asymmetric overlap it fills in the picture.

What overlap does not prove

It is easy to over-read this measurement, so the limits are worth stating plainly.

Resale is not fraud. A reseller that selects exits for speed, offers better targeting or tooling, or answers support faster is selling something real. A reseller can even be the fastest product in a comparison, because the curation is doing work. The issue is paying a premium for resold inventory without knowing, or relying on it for independence it cannot provide.

Shared ASNs are not shared addresses. Two networks that both source from the same large ISP look similar in an ASN breakdown while sharing no actual addresses. ASN-level similarity is a hint; only address-level matching is evidence.

Neighbours are not matches. On IPv4, two exits in the same /24 are often different households. On IPv6, one household usually holds a whole /64, so exact address matching misses shared devices: compare IPv6 exits by /64 instead. Mixing the two rules either hides real overlap or invents it.

Direction can change with time. A network that starts as a reseller and builds its own supply shows falling overlap month by month. A wholesaler that adds a new upstream shows a sudden jump. A single window is a snapshot; the trend says more.

What a second provider really buys you

Once you know the overlaps, you can work out what a multi-provider setup is worth. For two providers, the addresses you can actually reach are:

reachable exits = A + B − shared

With three, add all three, subtract each pairwise overlap, and add back the addresses all three share. In practice the useful number is simpler: for each candidate, how many exits would it add that you cannot already reach? That is its exclusive share of your setup, not its pool size, and it is the number to weigh against its price.

For redundancy, the question is narrower still. What matters is not total overlap but overlap on the targets where you need the fallback. Two networks can be independent on easy sites and share most of their working exits on a hard one, because only a small part of each pool gets through there.

How to use it when buying

Check overlap before you choose a fallback. If you are adding a second provider for redundancy, its overlap with your primary on your critical targets is the most important number in the comparison. It matters more than either network’s success rate.

Discount pool claims by exclusivity. The useful figure is not the advertised pool, nor even the measured one. It is the measured pool multiplied by its exclusive share: the addresses you genuinely cannot reach any other way.

Ask about supply in writing. A provider that owns its supply will usually say so plainly. A vague answer to “do you operate the exits you sell, in which countries?” is itself a data point.

Treat a resale signature as a business risk, not just a technical one. A reseller’s product changes when its upstream’s terms, prices or quality change, and you have no relationship with the upstream. For a prototype that hardly matters. For anything with a deadline attached, it belongs in your risk register.

Doing it yourself

If you have accounts with two or more providers, this is a day’s work:

  1. Run both at the same time. Send the same requests through each provider in the same window. Pools rotate constantly, so sampling one on Monday and the other on Thursday tells you nothing.
  2. Record what the target saw. Log the exit address from a target-side check, such as an IP echo service, on every request, not what the provider’s dashboard reports. Dashboards can be cached, wrong or describing the gateway.
  3. Take two samples per provider. Split each provider’s traffic into two halves of equal size, so you can measure its overlap with itself.
  4. Sample enough. Aim for tens of thousands of distinct exits per provider. Below a few thousand, every overlap looks small and the comparison with self-overlap becomes noise.
  5. Normalise addresses. Compare IPv4 exactly and IPv6 by /64.
  6. Compute both directions. Shared count as a share of each pool, then the relative overlap against each pool’s self-overlap.
  7. Repeat next month. The change between windows is often the most telling result.

How Atlas measures it

Atlas runs this measurement for every pair of networks it tests and publishes it on the proxy pool overlap page, including the direction of each overlap and each network’s exclusive share. Every network is tested in the same window with the same requests, and the exit addresses each one handed out are intersected pair by pair. The methodology describes exactly how it is done and where its limits are.

Few comparisons publish anything like it. That is odd, given it is one of the few things in this market that can be checked rather than claimed.

Frequently asked questions

How can I tell whether two proxy providers share the same IP pool?

Send the same requests through both providers in the same window and log the exit address the target saw on every request. Intersect the two sets, and compare the result with each provider's overlap with itself, because small samples make every overlap look low. Shared addresses show shared supply, not common ownership.

Does using two proxy providers guarantee better IP diversity?

No. Two providers may draw on overlapping networks or the same upstream suppliers. To judge diversity, compare the addresses each one actually hands out, where they are and which networks they belong to. How independent the pools are matters more than how many providers you pay.

Why do different proxy providers sometimes return the same IP addresses?

Residential supply is built by a handful of operators, and wholesalers and resellers sell that capacity on under their own brands. One brand can also mix its own supply with bought-in supply. So two brands can hand out the same household address even when their dashboards, plans and prices look nothing alike.

Get started

Test a second pool before you rely on it

Order Shifter's residential network with a 3-day money-back guarantee and compare its exit addresses with the provider you already use, before you count it as a second source.

3-day money-back guarantee. Cancel anytime.