ISP vs residential proxies is a choice between holding a few addresses and borrowing many. With ISP proxies you hold fixed addresses that a consumer broadband provider owns and leases out, running on a server rather than in a home: each is dedicated to you, stays the same until you release it, and is billed per address per month. A residential proxy sends each request, or each sticky session, through the connection of a household that opted in to a pool: the address changes, no address is yours, and the bill is per gigabyte. Which one fits depends on how long your workload has to keep one identity. Which one costs less depends on two numbers you can measure: how many distinct addresses the job needs, and how much traffic it moves.
Below: the naming overlap, a side-by-side table, how sessions and identity behave on each, the billing difference with a worked example you can rerun with your own figures, where each kind of address comes from, and a decision procedure.
The short answer
Choose ISP when one identity must keep one address for longer than an hour, when a partner has to allowlist your egress, or when you need to choose the country or the consumer ISP at checkout. Choose residential when the job needs many distinct exits relative to its traffic, or the location spread of a large pool. When both would do the job, compare the bills: at the rates published on 16 October 2026, ISP costs less once each address you would hold carries more than about 0.6 GB a month, and residential costs less below that.
Two names for one product, one name for two
ISP proxies are sold under two names. "ISP proxy" describes where the address comes from: a lease from a consumer internet provider. "Static residential proxy" describes how a target may classify it: as residential, because the address belongs to a broadband provider's space. Both refer to an address leased from the ISP and terminated on the proxy vendor's server, which is why it does not change and why its throughput is a server's rather than a home line's. Our static residential proxies guide covers how the leases work.
Residential proxies are a different product. Traffic exits through home devices whose owners opted in, typically via a partner app, so the exit really is a home connection. The pool is large and changes constantly as devices come online and go offline.
The overlap runs one step further: a few vendors use "static residential" for a long sticky session on a real household device, which ends when that device goes offline. When a quote says "static residential", ask whether the address is leased and server-hosted or sits on a peer's connection. The answer changes uptime, billing and the sourcing questions you need to ask.
ISP vs residential proxies side by side
The table compares the two lines as ProxyForge publishes them on the ISP proxies and residential proxies product pages.
| ISP proxies | Residential proxies | |
|---|---|---|
| What you buy | Specific addresses, held for a term | Traffic through a pool; no address is yours |
| Where traffic leaves | A server on our racks | A peer's household connection |
| How long an address lasts | Until you release it or it is replaced | One request, or a sticky session of up to 60 minutes |
| Distinct exits | The number you order | Whatever the rotation reaches, in a pool of 38M+ addresses |
| Throughput | 100 Mbps per address | Limited by each peer's connection |
| Billing unit | Per address, per month | Per GB, rounded to the MB |
| Smallest order | 1 address | 1 GB |
| Country choice | At checkout, with the consumer ISP | Not at checkout today |
| Allowlistable by a partner | Yes | No |
| Flagged address | Replaced within 4 hours | Not applicable: the next request takes another exit |
| Supply | Leased from ISPs under commercial contract | Peers who opted in and are compensated |
Session and identity behavior
On an ISP address, identity is something you hold. The same address answers every request for as long as you keep it, so a signed-in account, an allowlisted integration or a long multi-step flow sees one consistent origin for months. The price of that consistency is history: the address builds a record with each target. A good record helps; a block stays until the address is replaced, which we do within 4 hours for a flagged ISP address, and the replacement is a different address with its own record. Because a few addresses carry all the traffic, the target's per-address rate limit sets how many you need. By default no more than 64 addresses in a batch share one /24, so a target that limits per block does not see the batch as one source.
On a residential pool, identity is something you borrow. A rotating session takes a different household exit on each request, so per-address limits rarely bind and no history builds up on your side. A sticky session holds one exit for up to 60 minutes and ends sooner if the peer's device goes offline; rotating vs sticky proxies covers how to design flows around that window. Anything that has to appear as one returning visitor across days, such as an account signed in daily, will arrive from a different household each time.
To sort a workload, ask how long one identity must keep one address:
- Never the same twice: residential rotation. Breadth is the point.
- Minutes, within one flow: either. A residential sticky session covers it if the flow finishes inside the session window.
- Hours to months: ISP. A peer-hosted session is not built to last that long.
The billing difference, worked through
The two lines use different meters. ISP charges for the addresses you hold, whatever traffic they carry. Residential charges for the traffic, whatever addresses it crosses. Both ladders step down with order size, and an order is billed at the rate of the largest published rung at or below its quantity.
- ISP monthly cost = addresses held × the per-address rate for that many addresses × (1 − any term discount)
- Residential monthly cost = gigabytes ordered × the per-GB rate for that order size
Take a job that keeps 10 identities and moves 100 GB a month. At the rates published on 16 October 2026, with residential bought as one order a month and ISP on a 30-day term:
- ISP: 10 addresses earn the 10-address rate of $2.66 per address per month. 10 × $2.66 = $26.60. Traffic is unmetered at 100 Mbps per address, so the 100 GB adds nothing. We may step in if usage interferes with other customers.
- Residential: 100 GB earns the 100-GB rate of $2.65 per GB. 100 × $2.65 = $265.00.
- Crossover for 10 addresses: $26.60 buys about 5.8 GB of residential traffic at the 5-GB rate of $4.60 ($26.60 ÷ $4.60). Below about 6 GB a month in total, the residential order costs less; above it, the ten ISP addresses do.
Repeating the calculation along the ISP ladder, at the same published rates, gives the traffic at which the two bills meet:
| ISP addresses held | ISP rate per address | ISP cost, 30 days | Residential GB for the same money | GB per address |
|---|---|---|---|---|
| 1 | $3.20 | $3.20 | Below the 1-GB minimum ($5.50) | n/a |
| 10 | $2.66 | $26.60 | 5.8 at $4.60 | 0.58 |
| 25 | $2.40 | $60.00 | 15.2 at $3.95 | 0.61 |
| 50 | $2.14 | $107.00 | 31.5 at $3.40 | 0.63 |
| 100 | $1.89 | $189.00 | 63.0 at $3.00 | 0.63 |
| 250 | $1.70 | $425.00 | 160.4 at $2.65 | 0.64 |
| 500 | $1.50 | $750.00 | 326.1 at $2.30 | 0.65 |
Across the ladder, the bills meet at roughly 0.6 GB per address per month. To place your own job, divide the traffic it moves by the number of addresses it would need on ISP. Above about 0.6 GB per address, ISP costs less; below it, residential does. A crawl that needs 250 distinct addresses to stay under per-address limits but moves 50 GB a month sits at 0.2 GB per address: $425.00 on ISP against $150.00 for a 50-GB residential order at $3.00.
Three adjustments move the answer:
- Term discounts apply to ISP only. A longer term chosen at checkout takes up to 20% off the per-address rate, which moves the crossover lower.
- A residential order just below a rung can cost more than the rung. 9 GB at the 5-GB rate costs $41.40; 10 GB at the 10-GB rate costs $39.50. Check the next rung before ordering.
- Gigabytes run higher than estimates. Retries, headless browsers and uncompressed responses add to the residential bill and nothing to the ISP bill; proxy pricing per GB vs per IP shows how to estimate them.
Recompute with the current ladders on the ISP pricing and residential pricing pages before you decide; the figures above are a dated example, not a quote. And the comparison only matters where both lines would do the job: a rotating pool cannot hold an allowlisted address, and a handful of static addresses cannot spread a crawl across thousands of exits.
Running both lines on one workload
Many jobs have a breadth part and an identity part. A price-monitoring job might discover product URLs across a whole catalogue, which wants many exits and little traffic per exit, and then sign in to a handful of accounts to read prices shown only to members, which wants one stable address per account. Splitting the job puts each part on the meter that suits it: the discovery crawl on residential, billed by the gigabytes it moves, and the accounts on ISP, one address each, billed per address.
The split also protects the account addresses. Their history with the target is built only by the account traffic, because the crawl's volume never passes through them. Both lines draw on the same prepaid wallet and speak HTTP, HTTPS and SOCKS5, so the split is a routing rule in your client rather than a second integration. Size each side separately with the formulas above: the crawl by its gigabytes, the accounts by their count.
Where the addresses come from
ISP addresses are leased directly from consumer internet providers under commercial contracts and hosted on our own racks; no peer device is involved, so their supply rests on a commercial contract. Residential addresses reach us through partner applications: the person using each device is shown a plain-language notice, must act to accept it, is compensated, and can reverse the choice from the application's own settings. Both channels are named in our published supply-chain policy, and an independent assessor reviews our acquisition records twice a year. The policy is published in full on the sourcing page, and what IP provenance means explains what a procurement review asks for each type.
A decision procedure for ISP vs residential proxies
- Will a third party allowlist your egress addresses? Choose ISP. A pool cannot be allowlisted.
- Must one identity keep one address beyond 60 minutes? Choose ISP. A residential sticky session ends at 60 minutes, or sooner if the peer drops.
- Must the order name a country, or the consumer ISP? Choose ISP. Residential coverage spans 190+ countries, but country selection is not offered at checkout today.
- Does the job need breadth, meaning many distinct exits relative to its traffic or the location spread of a large pool? Choose residential.
- If both would work, compute both bills with the formulas above, your own address count and monthly gigabytes, and the ladders on the day you buy.
- Ask whether a hosting address would do. If the target accepts hosting ranges, datacenter proxies sit on the same per-address ladder as ISP, with lower latency and a shorter replacement window.
- Test before you commit. ISP includes a paid trial before you commit to a term; on residential, a small order serves the same purpose, since there is no monthly minimum. Judge both on content returned rather than status codes, and check a sample of ISP addresses with the proxy checker, and treat a registry lookup as one more piece of evidence rather than proof.