If you plan to buy ISP proxies, the label on the product page tells you the least. What decides whether an order works is a short list of facts a vendor can state in writing: where the addresses come from, whether they are exclusive to you, how many you must buy, where they sit, how fast they run, what happens when one is flagged, what you pay and how you can try them first. This guide gives eight checks in the order they tend to matter. Each one has the question to ask and what ProxyForge publishes on that point, so you can compare any vendor's answer with ours and with each other.
The checks assume the product you are buying is a static, dedicated IPv4 address leased from a consumer provider and hosted by the vendor. Our guide to static residential proxies, the same product under its other name, explains how the supply works.
1. Where the addresses come from
What to ask. How the vendor obtains the addresses and under what kind of agreement. Vendors arrange ISP leases differently: an ISP may be the registered holder, the announcing network, both or neither, depending on the lease. A vendor should be able to describe its sourcing in writing before you order.
Do not stop at the answer. Run the registry and routing lookups on a sample of the addresses you are given and keep the output. The records take a few minutes to read, and how to tell if a proxy is ISP or datacenter walks through the commands and where they disagree. Because arrangements differ, a lookup is evidence to weigh, not proof of the product category; a result that surprises you is a question for the vendor, not a verdict.
What ProxyForge publishes. ISP proxies are static IPv4 addresses leased from major consumer ISPs, including AT&T, Comcast and Verizon in the United States, and hosted on our own racks. Provenance rests on the commercial lease, not on peer devices, and our public supply-chain policy lists direct ISP leases as a permitted channel. The sourcing page sets out the policy and the twice-yearly independent audit of it. The ISP line is the one place where you can choose the provider at checkout, alongside the country.
2. Whether the address is yours alone
What to ask. Is each address assigned to one customer for the whole term, or can it be shared? What was it used for before, and does the vendor reuse addresses it has taken back? A shared address carries someone else's behaviour into your results, and a recycled one can carry history. The trade-offs are in our guide to dedicated vs shared proxies.
What ProxyForge publishes. The assignment is static and dedicated to one customer. No other customer's traffic runs through the address, and it does not rotate while you hold it. We do not publish a statement about an address's history before you received it, so that is a question to put to us in writing like any other.
3. How many addresses you have to buy
What to ask. What is the smallest order? Some vendors set a floor of several addresses or more, which is fine for a large deployment and wasteful for a single account or a test. A minimum also raises what it costs to find out that the product does not suit you.
What ProxyForge publishes. The minimum order is one address, and there is no monthly minimum beyond the address itself. You fund a wallet in the dashboard, place orders when you need them, and nothing is billed between orders.
4. Where the addresses are and how they are spread
What to ask. Three separate questions: which countries can you choose at checkout, can you choose the ISP, and how many distinct /24 subnets will your allocation span? Targets often apply reputation and rate limits to a whole /24, so fifty addresses in one /24 can be throttled together, while the same fifty across many /24s cannot. A vendor that only gives you "whatever is free" has made the decision for you.
You can measure the spread once you hold the addresses. This counts addresses per /24 from a file with one address per line:
cut -d. -f1-3 addresses.txt | sort | uniq -c | sort -rn
What ProxyForge publishes. Country selection at checkout across the countries listed on the product page, and provider selection on the ISP line. The default cap is 64 addresses per /24, and tighter spread is available on request.
5. Speed, bandwidth and protocols
What to ask. What is the port speed per address, is traffic metered or capped, and what latency should you expect to the targets you use? "Unlimited" is not an answer without a port speed beside it. Ask which protocols the gateway accepts and how you authenticate, because that decides whether your existing client works unchanged.
Then measure. A published median is a median across all customers and all targets. The number that matters is the one from your client to your target, so time it on the trial addresses with the same tool you will run in production.
What ProxyForge publishes. Each address runs at 100 Mbps and is unmetered at that port speed, so the monthly rate covers the address and not the traffic through it. We may step in if usage interferes with other customers. Median latency is published as 41 ms to regional targets. The gateway accepts HTTP, HTTPS and SOCKS5, with username and password or an IP allowlist for authentication.
6. What happens when an address is flagged
What to ask. Get three answers in writing. First, how quickly a flagged address is replaced. Second, what you are credited for the time it was unusable. Third, whether the replacement comes from a different /24 or the same one, since a replacement from the range that just got flagged may inherit the problem. Also ask how much notice you get when the vendor replaces an address on its own initiative, which matters if a partner has allowlisted it.
A guarantee that states no time and no credit is an intention. The service level in a contract is a different thing.
What ProxyForge publishes. A flagged ISP address is replaced within 4 hours. A service credit applies pro-rata plus 20% for any address that falls below target, and monthly address availability is published at 99.99%. Support has a named engineer with a 1-hour first response on this line. See the ISP pricing page for the service terms alongside the rates.
7. What the price actually is
What to ask. The price per address at your quantity, the term it applies to, and how a longer term changes it. Ask whether the rate on the page is the rate you will be billed or an entry price that applies to a quantity you will not buy. Entry prices are real, but they describe the smallest rung of the ladder.
A worked example shows why the quantity matters. These figures are computed from the published ladder at the rates published on 16 October 2026, for a 30-day term:
| Order | Rate per address | Cost for 30 days |
|---|---|---|
| 25 addresses | 2.40 USD | 60.00 USD |
| 100 addresses | 1.89 USD | 189.00 USD |
The second order has four times the addresses for a little over three times the cost, because the per-address rate steps down with quantity. A longer term takes a further discount on top of the volume rate. Check the ISP pricing page for the current ladder before you plan around these numbers, and compare per-address billing with per-gigabyte billing for your own traffic using our guide to proxy pricing per GB vs per IP.
What ProxyForge publishes. The full ladder, by quantity and by term, is on the pricing page with no rate reserved for a sales call. Billing is per address per month. Volume customers can ask for invoicing instead of a wallet.
8. Whether you can try it first, and how you leave
What to ask. Is there a trial, is it free or paid, and does it use the same addresses and gateway as a full order? A trial is only useful if it runs under the real conditions. Also ask what ending the arrangement involves: whether unused days are credited if you release an address early, and what you lose when a term ends.
Use the trial for the checks above. Run the registry lookups, count the /24 spread, time requests to your targets, and record the results. Those become the baseline you hold the vendor to.
What ProxyForge publishes. ISP includes a paid trial before you commit to a longer term, and you can start one from the dashboard or by asking sales. There is no monthly minimum, so a small first order is also a low-commitment test. If you are moving from another provider, the migration service maps your current endpoints and targeting onto our gateway and keeps both running in parallel until your numbers agree.
The order to run the checks in
The eight checks fall into three passes, and doing them in this order avoids paying for information you could have had for free.
- On paper, before you create an account. Checks 3, 6 and 7 are published terms: order size, replacement and price. A vendor that has not published them is asking you to negotiate what the competition prints. Rule it out here.
- With a small trial order. Checks 1, 2, 4 and 5 can only be confirmed on addresses you hold. Run the registry lookups on every one, count the /24 spread, and time requests to your real targets at the hours you will run.
- Before scaling. Check 8 and your own targets decide it. If the trial addresses behave on the targets that matter, scale the order, and re-run the lookups on the new addresses, since a larger allocation can span different ranges.
If a check fails in pass two, send the vendor the evidence and ask for the explanation in writing before you spend more.
Ask any vendor to put these in writing
Send the same list to every vendor you are comparing and keep the replies. The table gives the wording to ask for, and where ProxyForge publishes its answer.
| Check | Ask for in writing | ProxyForge answer is published on |
|---|---|---|
| Sourcing | How the addresses are obtained and under what kind of agreement | ISP proxies and sourcing describe the lease |
| Exclusivity | That each address is dedicated to you for the term, and what the vendor does with returned addresses | ISP proxies |
| Order size | The smallest order, and any monthly minimum | Pricing |
| Location and spread | Countries and provider you can choose, and the maximum addresses per /24 | ISP proxies |
| Speed and protocols | Port speed per address, whether traffic is metered, and the accepted protocols | ISP proxies |
| Replacement | Time to replace, credit for downtime, and whether the replacement leaves the flagged /24 | Pricing |
| Price | The rate at your quantity and term, and the term discounts | Pricing |
| Trial and exit | What the trial includes, and how an early release is credited | Pricing |
A vendor that answers all eight plainly has made your comparison easy. One that avoids three of them has told you how it will handle the first dispute. If datacenter addresses would serve the workload instead, the same questions apply to datacenter proxies and their pricing, where the ranges are registered to us and announced from our own ASN.