Are Residential Proxies Really Residential? How to Measure Any Pool

Every residential proxy vendor publishes a pool size. None of them publishes a pool composition — how much of that pool is actually a home broadband line, and how much is a server in a data centre wearing a residential label. That second number is the one that decides whether the residential premium you are paying buys anything at all, and it is measurable in an afternoon with free tools.

This page is both the method and the result: the exact rule, the exact commands, the sample size fixed in advance, and whatever the run produced — published as measured.

What a "residential" label actually promises

A residential proxy is sold on the premise that its exit address belongs to a consumer internet connection, so the traffic looks like an ordinary person browsing rather than a machine in a rack. That is the entire value proposition, and it is why residential bandwidth costs multiples of what datacenter bandwidth costs — the difference between the two is explained in residential vs datacenter proxies, and the underlying mechanics in what is a residential proxy.

What the label does not promise is that every address in the pool satisfies that description. A pool is assembled — from SDK partnerships, from reward apps, from resellers, from other proxy networks — and the further an address sits from the vendor's own supply chain, the less the vendor can say about it. Our reasoning on why that supply chain deserves scrutiny is in ethical proxy sourcing.

So there is a gap between the label and the addresses, the gap is a number, and the number is checkable. Nobody checks it, which is the only reason this page is interesting.

The method, fixed before the run

Four decisions define this measurement. All four were made before any data was collected, and all four are published so the run can be disputed rather than merely believed.

1. Only free, redistributable sources

Origin networks come from Team Cymru's public IP-to-ASN service and network types from PeeringDB, which is CC BY — free to use and republish with attribution, which this page gives.

The obvious easier option was rejected on inspection. ip-api.com returns a direct hosting boolean, which would have been richer signal than anything reconstructed from network names, but its terms restrict the free tier to non-commercial use in a non-commercial environment. This site monetises, so publishing figures derived from it would breach those terms — the same category of shortcut the piece criticises vendors for. Its free tier is also HTTP-only, and a measurement whose entire value is credibility cannot source its inputs over a channel any intermediary can rewrite.

That rejection is not free of cost, and it is the main reason the headline figure is a lower bound rather than an exact share.

2. A conservative classification rule

An address is counted as hosting when the registered name of its origin network contains one of a published list of terms — hosting, cloud, datacenter, colocation, hetzner, digitalocean, vps and similar. An address is counted as consumer when PeeringDB records its network as Cable/DSL/ISP. Everything else is unclassified and stays in the denominator.

Two properties of that rule matter more than the list itself. Hosting wins when both signals fire, because the conservative direction is to not under-report servers in a pool sold as residential. And addresses whose network could not be resolved at all are kept in the denominator rather than dropped — quietly shrinking the denominator would inflate whichever share the reader cares about, which is precisely the move this measurement exists to check somebody else for.

3. A sample size chosen in advance

1,087 unique exit IPs, which is the size that gives a plus-or-minus two percentage point margin at 95% confidence for a proportion near 13%. The measuring script refuses to start without an explicit sample size and prints the full method before the first request, so the number cannot be quietly tuned after the data arrives.

4. Every figure carries its interval

Shares are reported with a 95% Wilson score interval rather than a bare percentage. Wilson rather than the normal approximation because at proportions near zero the normal interval runs below zero and reports a lower bound that cannot exist. A pool statistic without an interval invites the reader to treat it as exact, which is the error the whole piece is arguing against.

Results

The table carries the most recent run for each pool. Every run ever published, superseded ones included, stays in the JSON at /data/pool-composition.json, licensed CC BY 4.0, alongside the sample size, the interval, the exact classification rule that run used, and a per-network tally that lets anyone re-classify it without re-sampling the pool.

The 17 September 2026 run: DataImpulse, US residential

Setup. DataImpulse's standard residential pool, targeted to the United States in the dashboard, rotating session, default targeting, anonymity filter off, no ASN exclusions — the pool a customer gets without touching a setting. 1,120 proxied requests yielded the 1,087 unique exit addresses fixed in advance. They resolved to 131 origin networks, all classified against a single PeeringDB export of 35,393 networks fetched at 06:12 UTC the same day. The run consumed roughly 1 MB of evaluation traffic that DataImpulse provided free of charge, which is disclosed here because it is the kind of thing that should be.

Hosting: at least 5.1%. 55 of the 1,087 exits came from networks whose registered name says data centre — a 95% interval of 3.9% to 6.5%. The list is short and specific: 45 of the 55 were Leaseweb USA, across eight of its sites (Dallas, Washington, Los Angeles, Miami, Phoenix, Chicago, New York and Seattle); nine were Colocation America; one was Dedicated.com. Read as a lower bound, that is one exit in twenty from a hosting rack, in a pool sold and priced as residential.

Consumer ISPs: 43.0%. 467 exits sat on networks PeeringDB registers as cable, DSL or ISP access — Comcast (109), RCN (98), Charter (81 across six registrations), T-Mobile (40), Cox (28) and Windstream (20) lead. Those counts are scoped to this bucket: Charter holds four further registrations, 22 exits, that PeeringDB gives no type at all, and they sit in the unclassified section below. These are the addresses the residential label promises.

Unclassified: 52.0%, and this time it is a fact about the pool. An earlier attempt at this run had to be discarded when PeeringDB rate-limited the per-network lookups halfway through, which would have made "unclassified" a measurement of the lookup service. This run judged every network against one export, so the 565 unclassified exits decompose cleanly:

• 334 (30.7%) are on networks PeeringDB has no record or no type for. Four operators account for 215 of them — DynaNode LLC (83), Rocks Computer Services (77), eSited Solutions (28) and Flash Edge Information Technology (27). None is a name that appears on a consumer broadband bill; none has a name the published rule recognises as hosting either, so the rule does not count them. Between them they are one exit in five. What they are is the single largest open question in this result.
• 194 (17.8%) are registered as NSP — network service providers such as Verizon Business (61), AT&T (57), Frontier (25) and CenturyLink (9). These carriers also sell consumer lines, so this bucket most likely leans residential; the rule does not assume so.
• 32 (2.9%) are registered as Content — Enzu, Hype Enterprises, Latitude.sh and Psychz Networks among them, operators that are hosting businesses in practice but whose names do not say so.
• 5 (0.5%) fall in one-off registry types — Educational/Research (2), Non-Profit (1), Network Services (1) and Enterprise (1). Too few to move the result, listed so the four groups add up to 565.

Put those pieces together and the true hosting share of this sample sits somewhere between the 5.1% the rule can prove and roughly 8% if the Content bucket is counted, while the consumer share sits between 43% and roughly 61% if the NSP carriers are. The two operators that decide where in that range the truth lies are DynaNode and Rocks Computer Services, 160 exits between them. The classification rule was not amended after these names appeared: adding them now would be choosing the answer after seeing the data, which is the practice this page exists to refuse.

What this does and does not say about Proxyway's figure. Proxyway's April 2026 benchmark put roughly one US exit in eight on a server using IP2Location's usage-type database. This run cannot confirm or contradict that number — different input, different rule, different quantity — and it does not try to. What can be said is that a free, conservative, name-only rule already finds one exit in twenty from named data centres, and that the direction of the two findings is the same.

Vendor response, 19 September 2026. These figures were sent to DataImpulse before publication, as promised; the company had previously told us in writing that it does not measure or publish a composition figure of its own. After checking the four unexplained networks with its technical team, it replied:

• DynaNode LLC and eSited Solutions are datacenter providers, in DataImpulse's own words, and the company says it has already asked its technical team to exclude both from the residential pool.
• Rocks Computer Services and Flash Edge Information Technology it cannot confirm as datacenter providers on the information it has, and it says they will remain in the pool.

Add the two networks the vendor identifies to the 55 exits the rule already counts, and at least 166 of the 1,087 exits — 15.3%, with a 95% interval of 13.3% to 17.5% — came from confirmed data centres. DataImpulse did that arithmetic itself and asked for the higher number to be the one quoted: "We'd rather have that number accurate than understated." Vendors rarely argue their own figures upward, and it is credited here. On 29 September DataImpulse confirmed three further operators in this sample as datacenter providers, which restates this run at 17.3%; the correction is in the 26 September section below.

The published measurement does not change. The rule still reports 5.1%, because it was fixed before the run and the vendor's confirmation is evidence the rule never saw; 15.3% sits beside it as a separate, attributed figure rather than being folded into it. It is above Proxyway's roughly 13%, though the two still come from different methods. What remains open is 104 exits (9.6%) on Rocks Computer Services and Flash Edge: if both were also data centres, the confirmed share of this sample would reach 24.8%, and nothing here says that they are.

DataImpulse offered fresh evaluation traffic for a retest once the exclusions were complete, and signalled completion on 22 September. That run is reported below, as a second dated run beside this one rather than in place of it.

The 22 September 2026 run: after the cleanup

Setup. Identical to the first run by design — the same standard residential pool, the same US targeting, the same rotating session and default settings, the same pre-registered sample size (a fresh draw of 1,087 unique exits from 1,120 proxied requests), and the same frozen classification rule, unchanged since before either run. They resolved to 127 origin networks, judged against a PeeringDB export of 35,441 networks generated at 22:50 UTC and fetched at 23:25 UTC the same day. DataImpulse supplied the evaluation traffic again, and asked that the run wait until it signalled its cleanup was finished. That was a reasonable request and it was granted, but it shapes what the result can mean, which the last paragraph of this section spells out.

Hosting by the rule: 0.7%. 8 of the 1,087 exits, a 95% interval of 0.4% to 1.5%, down from 5.1%. All eight were Colocation America. Leaseweb, which supplied 45 exits across eight of its US sites in the first run, supplied none at all; nor did Dedicated.com.

Consumer ISPs: 55.7%, unclassified 43.6%. 605 exits now sit on networks PeeringDB registers as cable, DSL or ISP access, up from 467 — Comcast (174), RCN (112), T-Mobile (49) and Cox (42) lead. The unclassified bucket fell from 565 exits to 474.

The four unexplained networks, five days on. eSited Solutions went from 28 exits to none, and Rocks Computer Services from 77 to none. Flash Edge fell from 27 to 10. And DynaNode LLC, the one network of the four that DataImpulse had positively identified as a datacenter provider and asked its team to exclude, went from 83 exits to 79: effectively unchanged, and still the fourth-largest source of addresses in the sample. Rocks going to zero read at the time as more than DataImpulse had committed to, since on 19 September it said it could not confirm Rocks as a data centre and that the network would stay. It was not: the company established a week later that Rocks had been removed by mistake, and that correction is printed below.

Nothing moved into new racks. Not one network with five or more exits appears in this run that was absent from the first. That is worth checking explicitly, because the easiest way to pass a retest is not to remove traffic but to move it to networks with different names. The fall here is removal.

Vendor response, 24 September 2026 — since retracted. Asked why DynaNode was unchanged, DataImpulse first replied: "DynaNode LLC operates under two ASNs. One has been removed (contained around 6,500 IPs). The second ASN is still in the process of being excluded. That's why 79 exits remained in your retest." That account is no longer operative. It is left standing here rather than deleted, because a correction whose original the reader cannot see is not a correction.

Two parts of it could not be reconciled with the public record. AS400463 is the only ASN registered under the DynaNode name in public registry data, and no second DynaNode-named network had appeared in the 17 September sample either. Neither observation proved anything on its own — an ASN can be registered to a different entity, and 6,500 addresses can easily produce no exits in 1,087 draws — so both were put to DataImpulse rather than published as a finding.

Correction from DataImpulse, 25 September 2026. The company checked again and withdrew the statement: "There isn't a second DynaNode ASN. The network that had already been removed was Rocks Computer Services, LLC, but it was mistakenly marked internally as DynaNode. That also explains why Rocks dropped to zero in your retest while AS400463 remained present." AS400463 is the DynaNode ASN, and at the time of this run it was still in the pool.

That correction carries a consequence that does not flatter the result above. Rocks Computer Services was removed by mistake. DataImpulse said on 19 September that it could not confirm Rocks as a data centre and that the network would stay, and nothing since has changed that assessment: those 77 exits left the sample because of an internal mislabelling, not because anyone concluded they should. The pool got cleaner on that network for the wrong reason. Whether the exclusion is now deliberate is a question put to the company and not yet answered.

Counted the vendor's own way: 8.0%. Add DynaNode, which DataImpulse identifies as a data centre, to the 8 exits the rule counts, and 87 of the 1,087 exits — 8.0%, with a 95% interval of 6.5% to 9.8% — came from confirmed data centres, against 15.3% five days earlier. Flash Edge's remaining 10 exits would take that to 8.9% if it too turned out to be hosting, which nothing here establishes. As with the first run, the published measurement stays at what the rule reports: 0.7% is the figure the method produced, and 8.0% sits beside it as a separate, attributed figure. That figure is also superseded: the same 29 September confirmation restates this run at 10.1%, as the correction below sets out.

What this run cannot show. It was taken after the vendor knew exactly which networks were being checked and told us it had finished. It shows those networks are gone from this sample. It does not show that the pool is cleaner in general, and a pool can be clean for a week. Two further runs were planned on the same frozen rule and the same pre-registered sample size: one once DataImpulse confirmed that AS400463 was excluded, which is the run reported next, and one some weeks later on a date not announced in advance.

The 26 September 2026 run: the network the vendor named is gone

Setup. Unchanged again — the same standard residential pool, the same US targeting, the same rotating session and defaults, 1,087 unique exits drawn from 1,120 proxied requests, and the same frozen classification rule. They resolved to 125 origin networks, judged against a PeeringDB export of 35,475 networks generated at 01:01 UTC and fetched at 01:22 UTC the same day. DataImpulse confirmed on 25 September that AS400463 had been removed and invited the run; the evaluation traffic was again supplied by the company.

DynaNode: 79 exits to none. AS400463 did not appear once in 1,087 draws. The network DataImpulse identified as a datacenter provider on 19 September, and which was still the fourth-largest source of addresses in the sample three days later, is gone. Of the other three networks tracked since the first run, eSited Solutions and Rocks Computer Services remain absent, and Flash Edge, which DataImpulse said it could not confirm as a data centre, fell again, from 10 exits to 4.

Hosting by the rule: 0.7%. 8 of the 1,087 exits again, a 95% interval of 0.4% to 1.5% — seven Colocation America and one IHOR Hosting.

Correction, 29 September 2026: this was first published as a convergence, and it was not one. What stood here said that the rule's 0.7% and DataImpulse's own identification had arrived at the same number for the first time in the series. Three days later the company confirmed that Hype Enterprises, Enzu Inc and Psychz Networks — the operators the paragraph below listed as unexplained — are datacenter providers, and that it had excluded their ASNs. Those three supplied 28 exits in this sample. Counted the vendor's own way, this run is 36 of the 1,087 exits, 3.3%, with a 95% interval of 2.4% to 4.6%: not 0.7%, and not a convergence.

The same confirmation reaches backwards, because all three operators were present in every sample. Restated on what DataImpulse has now confirmed, the series reads:

• 17 September: 5.1% by the rule, 17.3% confirmed (188 of 1,087 exits), published at the time as 15.3%.
• 22 September: 0.7% by the rule, 10.1% confirmed (110 exits), published at the time as 8.0%.
• 26 September: 0.7% by the rule, 3.3% confirmed (36 exits), published at the time as 0.7%.

The direction is unchanged and the improvement is real. What changed is the distance between the two columns — and that is this article's own argument turned on itself. Between 22 and 26 September the rule never moved off 0.7%, while the confirmed share was 10.1% and then 3.3%. Nothing about the pool produced those revisions. The only variable was which networks had been put to the vendor by name. A lower bound is not an estimate, and every figure the rule produces should be read as the floor it is.

Consumer ISPs: 61.8%, unclassified 37.4%. Across the three runs the consumer share has moved 43.0% → 55.7% → 61.8%, and the unclassified share 52.0% → 43.6% → 37.4%. Comcast (172) and RCN (146) lead, then T-Mobile (60), Cox (52) and Cablevision (31); Charter accounts for 101 exits across its consumer-bucket registrations.

Nothing moved into new racks, again. One network with five or more exits appears here that was absent from the second run: Castro Telecom (AS33733, 5 exits), a second registration of an operator already in the sample, and not a hosting name. The stronger evidence is in the aggregate — the group of networks PeeringDB has no record or no type for, counted as in the first run without the exits the rule already counts as hosting, shrank from 234 exits to 140. Had the removed traffic simply been relabelled onto differently-named networks, that group would have grown.

What was still unexplained, and no longer is. 38 exits (3.5%) sat on operators PeeringDB registers as Content — Hype Enterprises (12), Enzu (11) and Psychz Networks (5) among them — hosting businesses whose registered names the rule does not recognise. Those three names were put to DataImpulse, which confirmed on 29 September that all three are datacenter providers and that it had excluded their ASNs. That is the confirmation the correction above is built on. Five smaller operators in the same bucket — Clouvider (3 exits), PEG TECH (2), H4Y Technologies (2), Railway (2) and Antbox Networks (1) — have been put to the company and are not yet answered.

What this run settles, and what it does not. It settles that AS400463 is out of the pool, which is what it was run to check. It does not settle that the pool stays this way, because every run in this series has been taken on a date the vendor knew in advance — and the run reported next, three days later, is the clearest demonstration of why that limit matters.

The 29 September 2026 run: the rule reads 0.1%, and that is the least useful figure in this article

Setup. Unchanged — the same pool, the same US targeting, the same defaults, 1,087 unique exits from 1,120 proxied requests, the same frozen rule. They resolved to 124 origin networks, judged against a PeeringDB export of 35,487 networks generated at 08:21 UTC and fetched at 08:51 UTC the same day. DataImpulse had said the same morning that it had excluded Hype Enterprises, Enzu and Psychz Networks, and invited the run.

The confirmed networks are gone, all of them. Hype Enterprises went from 12 exits to none, Enzu from 11 to none, Psychz Networks from 5 to none. DynaNode, eSited, Rocks Computer Services and Leaseweb stayed at zero. Every network DataImpulse has ever confirmed as a data centre is now absent from the sample.

Hosting by the rule: 0.1%. One exit of 1,087, a 95% interval of 0.02% to 0.5% — a single Colocation America address. That is the lowest figure this benchmark has produced, and taken alone it would say the pool is essentially clean.

Taken alone it would be wrong. Flash Edge Information Technology (AS219329) supplied 4 exits on 26 September and 75 on 29 September — a nineteen-fold increase in three days, and enough to make it the second-largest source of addresses in the sample. Flash Edge is the network DataImpulse told us on 19 September it could not confirm as a datacenter provider and would leave in the pool. PeeringDB holds no type for it, and the rule does not recognise its name, so the rule cannot see it. If Flash Edge is a data centre, this run is 76 of 1,087 exits — 7.0%, with a 95% interval of 5.6% to 8.7% — not 0.1%.

The buckets moved backwards, too. Consumer ISPs fell from 61.8% to 56.3% and unclassified rose from 37.4% back to 43.6%, reversing most of the previous run's gain. Two networks with five or more exits are new since the last run: Grande Communications (14 exits, registered as consumer cable) and GoCodeIT (7, untyped).

What we are and are not saying. We are not saying this was arranged. Removing four operators in ten days has to move that traffic somewhere, and a surviving supplier absorbing it is an ordinary consequence rather than evidence of intent. What we are saying is narrower and harder to argue with: the rule's headline figure improved while the share of the pool it cannot classify grew, and the growth landed almost entirely on the one network the vendor had been asked about and declined to classify. That is the exact failure mode a single pool statistic invites, which is why this page has never printed one on its own. DataImpulse's answer, and what the next run found, are in the 13 October section below.

The 13 October 2026 run: the rule finds nothing, and the third-largest network in the sample is unclassified

Setup. Unchanged — the same pool, the same US targeting, the same defaults, 1,087 unique exits from 1,120 proxied requests, the same frozen rule. They resolved to 116 origin networks, judged against a PeeringDB export of 35,608 networks generated at 09:41 UTC and fetched at 10:06 UTC the same day. On 13 October DataImpulse told us it had excluded the six remaining operators we had put to it — Clouvider, PEG TECH, H4Y Technologies, Railway, Antbox Networks and Datacamp — and invited the run.

The exclusions held, with one exception. Clouvider, PEG TECH, H4Y Technologies, Railway and Antbox Networks supplied no exits. Datacamp Limited (AS212238) supplied 5, down from 10; whether its exclusion is still propagating was put to the company, which answered on 25 October that after a manual review it classifies Datacamp Limited (AS212238) and GoCodeIT (AS835 / AS62513) as datacenter networks and is removing both from the residential pool. On 27 October the company told us both had been removed and invited a re-measurement, which is reported below. Hype Enterprises, Enzu, Psychz Networks, DynaNode, eSited, Rocks Computer Services and Leaseweb all stayed at zero.

Hosting by the rule: none. Not one of the 1,087 exits sat on a network whose registered name the rule recognises as hosting — a 95% interval of 0% to 0.4%, and the first run in which the rule found nothing at all. Taken alone, that would say the pool is clean.

Taken alone it would be wrong, again. Flash Edge Information Technology (AS219329) supplied 105 exits, 9.7% of the sample — 27, 10, 4, 75 and 105 across the five runs. It ranked third among 116 networks, ahead of AT&T (70 exits), Verizon (55), Frontier (53) and Cox (50); the largest other network PeeringDB holds no type for supplied 21. The rule cannot see it. If Flash Edge is a data centre, this run is 105 of 1,087 exits — 9.7%, with a 95% interval of 8.0% to 11.6% — not 0%. Consumer ISPs stood at 58.0% and unclassified exits at 42.0%, little changed from 29 September, and no network with five or more exits is new since that run.

What DataImpulse told us about Flash Edge. Asked why the network grew nineteen-fold between 26 and 29 September, the company wrote on 13 October that its residential IPs are sourced through TraffMonetizer, "our own SDK-based supply service", and that "An IP provider within that network began routing Flash Edge addresses into the pool, which is why the count jumped between 26 and 29 September. It was not arranged in response to the test." It said Flash Edge Information Technology LLC operates flashproxy.com, "which is itself a proxy service offering datacenter, mobile, residential, and premium residential proxies", does not cooperate with DataImpulse directly, and that "Our best assessment is that their IPs enter our pool through their end users participating in TraffMonetizer, but we cannot confirm that with certainty." On whether to exclude it: "Flash Edge runs a mixed network. If you need a binary answer for the article, the honest one is: we don't know yet whether these specific IPs are residential or datacenter, and we are still reviewing." On 25 October it answered that it buys no addresses from other providers and that all its residential supply, app and SDK alike, comes through TraffMonetizer clients, so the Flash Edge addresses entered through those clients; it cannot determine whether they are Flash Edge's own end users or Flash Edge itself making addresses from its network available, and it declined to disclose how many installs or partner IDs sit behind AS219329. It said TraffMonetizer checks the quality of each connection before passing it to DataImpulse and does not pay for connections that show problems; we print it as the vendor's statement. On 27 October it added that the check is not only about connection quality: each IP is also cross-checked against several IP-classification services, which decide whether it goes into the residential or the datacenter pool, and that it can share nothing more about who operates the clients behind AS219329. Flash Edge nevertheless reached the residential pool and the company says it still cannot classify it, so these checks do not settle whose device the connection runs on. The company has still not decided whether to exclude Flash Edge.

How DataImpulse finds data centres. Asked on 7 October how it identifies a data-centre network before anyone names it, the company described "an internal service that combines data from three sources: IP-API, DB-IP, and a proprietary system that checks IP addresses not only by location but also by proxy type", filtering IPs on that classification, and confirmed it pays for IP-API. It added: "The system exists and runs proactively, but as this series has shown, it did not catch everything that your method found." It also said Rocks Computer Services was judged a dedicated server operator by a manual review of the provider, its services and its infrastructure, and that no database classified it as an ISP. That is a filter, not a measurement: a filter shows what it removed, not what it missed.

What public registries add, and what they do not. Flash Edge's own site, flashproxy.com, sells datacenter, ISP, mobile and residential proxies; it says its residential IPs are "sourced through ISP partnerships", makes no statement about end-user consent, and does not mention TraffMonetizer. RIPE registry data show AS219329 was created on 29 June 2026 and announces 118 prefixes through a single upstream network, GoCodeIT Inc — a name that also appeared in our samples (27 exits on 17 September, 11 on 29 September, none on 13 October). Of the 108 prefixes whose registration could be resolved, 22 sit in address blocks registered under RapidSeedbox names, a seedbox and proxy hosting company. When the prefixes were first announced does not explain the jump: they appeared between July and October, the largest batch of 57 on 17 September. None of this classifies Flash Edge, and the rule does not use it. Two parts of DataImpulse's account remain to be reconciled: it gave two mechanisms — an IP provider within the network routing addresses in, and Flash Edge's own end users taking part in TraffMonetizer — and for a TraffMonetizer client to exit from an address that Flash Edge's network announces, its traffic would have to be pointed at a Flash Edge proxy. Both questions, and how many distinct installations account for these exits, have been put to the company.

The 27 October 2026 run: the rule finds two exits, and Flash Edge falls to 42

Setup. Unchanged — the same pool, the same US targeting, the same defaults, 1,087 unique exits from 1,120 proxied requests, the same frozen rule. They resolved to 107 origin networks, judged against a PeeringDB export of 35,752 networks generated at 03:03 UTC and fetched at 03:33 UTC the same day. DataImpulse told us on 27 October that Datacamp Limited (AS212238) and GoCodeIT (AS835 / AS62513) had been removed from the residential pool and said we could re-measure whenever it suited us; this run started within hours.

The removals held. Datacamp supplied no exits, down from 5. Neither GoCodeIT ASN supplied any, as on 13 October. Every other network DataImpulse has ever confirmed or removed — DynaNode, eSited, Rocks Computer Services, Hype Enterprises, Enzu, Psychz Networks, Leaseweb, Clouvider, PEG TECH, H4Y Technologies, Railway and Antbox Networks — also stayed at zero.

Hosting by the rule: two exits. 2 of 1,087 exits (0.18%, 95% interval 0.05% to 0.67%) sat on a network whose registered name the rule recognises as hosting: one on Dedicated.com (AS63018), which had also supplied one on 17 September, and one on Mullcloud LLC (AS397006), which no earlier run had seen. Neither has been put to DataImpulse. A single exit on a network is not a pattern, and we report it because the rule is frozen and the result is published whatever it says.

Flash Edge fell by more than half, and nothing we were told explains it. Flash Edge Information Technology (AS219329) supplied 42 exits, 3.9% of the sample (95% interval 2.9% to 5.2%) — 27, 10, 4, 75, 105 and 42 across the six runs. It ranked seventh among 107 networks, behind RCN (176 exits), Comcast (154), AT&T (64), Cox (64), Frontier (52) and Verizon (50), and ahead of Charter (38); the largest other network PeeringDB holds no type for supplied 26. The rule cannot see it. If Flash Edge is a data centre, this run is 44 of 1,087 exits — 4.0%, with a 95% interval of 3.0% to 5.4% — not 0.2%. DataImpulse's message of 27 October said nothing about Flash Edge, and its decision on whether to exclude it is still pending, so we do not know whether the fall reflects a change by DataImpulse, by TraffMonetizer or by Flash Edge itself. Consumer ISPs stood at 64.9% and unclassified exits at 35.0%, against 58.0% and 42.0% on 13 October, and no network with five or more exits is new since that run.

Flash Edge, on 28 October. DataImpulse wrote again after the run. Nothing has changed on its side: it has not excluded Flash Edge or adjusted it, and the addresses still enter the pool through TraffMonetizer, where, in its words, "someone is sharing IP addresses from this provider." It has not decided to exclude the network. That sentence does not say whose device the connection runs on, or how that person agreed to share it, so the question we raised on 13 October is unchanged.

Where the series stands. Six runs, one frozen rule, one sample size, every figure published with its interval. By the rule: 5.1%, 0.7%, 0.7%, 0.1%, 0.0%, 0.2%. Counted as the vendor has confirmed: 17.3%, 10.1%, 3.3%; for the last three runs, anywhere from 0.1% to 7.0% (29 September), from 0% to 9.7% (13 October) and from 0.2% to 4.0% (27 October), depending on a single review DataImpulse has not finished. The direction over the 40 days since the first run is genuinely down, and DataImpulse has removed every network it was shown and confirmed. What none of it establishes is durability, because the vendor chose or knew the date every time, and the largest open question has shrunk but not closed. The run that tests durability is the last one in this series: some weeks from now, on a date not announced in advance, from an account the company cannot identify, paid for by us rather than supplied as evaluation traffic. Same rule, same sample size, published beside these six whatever it says.

Run the check yourself, before you pay

The useful version of this article is the one you run against the pool you are about to buy. Rotate the endpoint, collect unique exit addresses, then ask Team Cymru what network each one came from. The lookup is a plain DNS query, so it needs no key and no account:

# 1. Get an exit IP through the proxy you are evaluating.
ip=$(curl -s --proxy "http://$PROXY_HOST:$PROXY_PORT" \
          --proxy-user "$PROXY_USER:$PROXY_PASS" https://api.ipify.org)

# 2. Ask Team Cymru which network announces it (reverse the octets).
reversed=$(echo "$ip" | awk -F. '{print $4"."$3"."$2"."$1}')
asn=$(dig +short TXT "$reversed.origin.asn.cymru.com" | awk -F'"' '{print $2}' | awk '{print $1}')

# 3. Resolve that ASN to the name its operator registered.
dig +short TXT "AS$asn.asn.cymru.com"

Repeat that a few hundred times and count how often the name that comes back reads like a data centre. If you would rather not build the loop, the full pre-registered run — sampling, classification, Wilson intervals and the JSON artifact — is one command in our open method:

pnpm run measure:pool -- --provider=dataimpulse --sample=1087

Two practical warnings. Send the proxy password over stdin or an environment variable, never as a command-line argument: arguments are readable in the process list of a shared machine for as long as the request runs, and this check issues thousands of them. And be polite to PeeringDB — it is a volunteer-run service that allows roughly one anonymous request an hour for its full export, so fetch the whole registry once, keep it on disk, and judge every address in a run against that one snapshot rather than looking networks up one at a time. That is also what makes the run reproducible: the artifact records which snapshot it used.

What a hosting share does and does not tell you

A high hosting share is not proof of dishonesty, and reading it that way is the fastest way to get the analysis wrong.

Some hosting-registered ranges are genuinely assigned to consumer connections. Some consumer ISPs operate their own hosting ranges under the same name. Some addresses are simply unclassifiable from public data, which is why unclassified is a published column and not a rounding error.

What a high hosting share does tell you is commercial. Datacenter bandwidth is cheap and residential bandwidth is not; a pool with a large server component is selling you cheap inventory at the expensive price. It is also operationally worse, because hosting ranges are the first thing a target site's anti-bot vendor buys a list of — which is the mechanism behind most of the blocks described in how to scrape without getting blocked. And if you are collecting data under a compliance obligation, "we could not account for one address in eight" is a sentence you do not want in an audit.

The inverse is equally worth stating: a low hosting share is evidence about composition, not about consent. A pool can be entirely consumer-ISP and still have been assembled from people who never meaningfully agreed to route strangers' traffic. Composition and sourcing are two different questions, and this measurement answers only the first.

Choosing a provider you can actually account for

Until composition figures are routine, the practical proxy for them is whether a vendor will document where its addresses come from and who it will sell to. That is observable today, from published pages rather than sales calls, and it is summarised across the market in the proxy market report. The short version:

And the vendor whose pool this benchmark was pointed at first:

Choosing between them on price rather than posture is a different exercise — that one is in best residential proxies and proxy pricing comparison.

Limitations

Stated plainly, because a method section that hides its weaknesses is marketing.

The hosting share is a lower bound, not an estimate of the true share. The rule only fires on self-describing network names, so every neutrally-named hosting operator is missed.

These figures are not comparable with any number derived from a commercial usage-type database, including the Proxyway benchmark cited above. Different inputs, different rule, different quantity. Presenting them as two readings of the same thing would be wrong even when they agree.

A pool is not static. A sample describes the addresses that pool handed out during one run, from one vantage point, at one moment. Composition can differ by target country, by time of day, and by how much you are spending.

And the sample says nothing about success rate, latency or consent. It is one property of one pool, measured carefully. Treat it as that and it is useful; treat it as a verdict on a vendor and it is not.

This is an independent measurement, not a vendor-supplied statistic, and not legal advice. If you re-run the method and get a different answer, we want the correction — that is what publishing the rule is for.