Pricing that requires a meeting
A small buyer should be able to see a per-GB or per-IP baseline without handing over an email address first. If a pricing page redirects to a contact form, the renewal rate, overage rate, and "fair use" terms are all going to arrive after the balance is already loaded.
Proxynade publishes current rates on the pricing page: Volume Residential at $0.89/GB, Premium Residential at $5.00/GB, billed per transferred byte. Static ISP proxies are priced per IP with unlimited bandwidth. No discovery call needed.
Pool-size banners are not testable
A number on a marketing page is not a claim you can verify. What you can do: run 50 requests, log the exit IP and ASN for each one, and count unique IPs. That is a small sample, but it is yours. Repeat with a narrow country filter and a broad one — the distribution changes, and ASN concentration becomes visible.
One annoying trial looked like a pool problem for half an hour because the same broadband ASN kept returning under a broad country filter. The actual bad rows were simpler: two curl repros returned 407 Proxy Authentication Required because the copied username had session_60s where the generated line used session-60s. Wrong separator, not a bad pool.
Log host, outcome, latency, and bytes per request during the trial. That data is what the Proxynade dashboard network log gives you on live traffic. If a vendor does not expose something equivalent, you cannot reconcile a dispute later.
Success-rate claims measure different things
A vendor may count a successful TCP connection as a success. The buyer probably wants: a complete response body, no target-side 403, no rate-limit 429, and no CAPTCHA after the page loads. Those are four different tests. Until the vendor says which one it measured, a percentage figure carries no information.
Run the measurement yourself against the specific target and workflow that matters. The number will be different from the marketing claim — that is normal. The useful number is the one from your own logs.
Provider bytes will not match app bytes
The proxy meter counts redirects, retries, failed response bodies, discarded assets, and prefetched files. Your app counts what it kept. Both meters are accurate; they measure different slices of the same traffic.
The gap is predictable once you know it is there. The Proxynade dashboard logs byte totals per request, which makes it possible to trace large gaps to specific hosts or retry storms. Without per-request byte data on the provider side, a billing dispute reduces to two numbers with no shared context.
| What the proxy counts | What your app counts |
|---|---|
| All bytes transferred, including retries and failed bodies | Bytes from kept responses only |
| Redirect hops | Final destination only |
| Prefetched assets your client discarded | Assets the page used |
| TLS overhead and CONNECT tunnels | Application-layer content |
No-KYC pools share exits with whoever paid
No-KYC panels feel low-friction until the same exits are shared with card testers and spam runs that got blocked elsewhere — and those blocks propagate through shared reputation lists like Spamhaus. Exit age and cleanliness are downstream of who uses the pool. Providers with no verification have no way to remove the accounts driving the blocks.
Residential, static ISP, and datacenter pools also degrade differently. A provider selling one vague "premium" bucket is asking the buyer to guess what failure mode is coming.
Connection details should be in the docs before purchase
Host, port, protocol, credential format, and sticky-session token format should all be in public documentation. A CI job should not need a support ticket to avoid a 407. If the expanded username scheme is not documented — the base user, plan token, optional country code, optional rotation window — that ambiguity produces silent misrouting that looks like a pool problem.
Also check the acceptable-use policy for the proxy type you are testing. Checkout flows, login endpoints, ticketing sites, and financial accounts are sometimes disallowed. Discovering a category restriction after loading quota is a waste of both the quota and the time spent debugging what looked like a target block.
Support signal before the trial ends
Search the vendor's public community — Discord, Telegram, forum — for 407, "refund," and your target category. Silence can mean stability. It can also mean nobody is staffing the queue. Recent unanswered threads are the signal worth weighing.
Send one routine support question during the pilot. Response time and accuracy on a non-urgent question is a reasonable proxy for what happens when the billing question arrives.
Buying proxy FAQ
What is the first thing to check before buying proxies? Check whether per-GB or per-IP pricing is published without a sales call. Opaque pricing means renewal and overage terms also arrive after the balance is loaded.
Why do provider bandwidth numbers not match my app? The proxy meter counts redirects, retries, failed bodies, and prefetched assets. Your app counts what it kept. Both numbers are accurate; they just measure different things.
How do I verify a pool-size claim? Run 50 requests, log the exit ASN and IP each time, and count unique IPs. That is a small sample, but it is concrete. A marketing number is not checkable; your own log is.
What does a 407 mean during a proxy trial? 407 Proxy Authentication Required means the proxy rejected the credentials. Check the username format, password, account balance, and whether you copied a token with the wrong separator.
What support signal should I check before buying? Search the vendor's public community (Discord, Telegram, or forum) for 407, "refund," and your target category. Recent unanswered threads in those searches are a signal worth weighing.
Pre-purchase checklist
- Per-GB or per-IP rate is published without a meeting.
- Connection string format is documented: host, port, credential format, sticky-session token.
- Per-request log data is available on the provider side for billing reconciliation.
- Pool type (residential, static ISP, datacenter) is explicit — not one vague tier.
- Acceptable-use policy covers your target category before the balance is loaded.
- 50-request trial with exit IPs logged before committing to volume.