The residential proxy market has no shortage of providers. Most of them make identical claims: millions of IPs, global coverage, 99.9% uptime, ethically sourced. Some of those claims are true. Many aren’t, and the ones that aren’t will cost you in ways that don’t show up until you’re running production workloads — degraded success rates, wrong geodata, silent failures, and support that disappears when something breaks at 2 AM.
Here’s what to actually verify before committing to a provider.
1. How the IPs Are Sourced
This is the question most buyers skip, and it’s the one that matters most — for legal, ethical, and practical reasons.
Residential IPs should come from users who have explicitly opted in to sharing their bandwidth through a partner application. They know their device is part of a proxy network. They’re compensated for it. This is the ethical baseline, and it’s also the legal one in most jurisdictions: using devices without owner consent to proxy traffic is unauthorized computer access in the US, UK, and EU regardless of what the provider’s terms say.
Beyond ethics, sourcing quality affects IP quality. Networks built on opt-in bandwidth tend to have healthier IPs — lower block rates, more diverse ASN distribution, more realistic traffic histories — because the devices are actively maintained by owners who care about their connection. Networks built on questionable sourcing tend to have IP pools full of addresses that have already been flagged by major platforms from previous abuse.
What to ask: Does the provider publish their sourcing methodology? Do they name their partner programs? Can they explain how IPs are removed from the pool when a user opts out? Vague answers (“ethically sourced” with no elaboration) are a red flag.
2. Pool Size and Geographic Distribution
Headline IP counts are the least reliable number in any provider’s marketing. “200 million IPs” means nothing if 180 million of them are concentrated in three countries, or if the pool hasn’t been refreshed in six months and a large share are blocklisted.
What matters in practice:
Effective pool size per target. If you’re scraping a specific domain at volume, the relevant number is how many unique IPs the provider can route through that domain without significant reuse. A 10M IP pool where you’re making 1M requests to the same domain has a 10% expected reuse rate — not ideal. A 100M pool at the same volume has 1%.
Country and city coverage. Country-level targeting is commodity. City-level targeting is where providers diverge. If your use case requires geo-accurate data — local SERP results, location-specific pricing, regional content — verify that city targeting actually resolves to IPs that geolocate correctly in third-party databases (MaxMind, IP2Location). Some providers claim city targeting but deliver IPs that resolve to the wrong city in the databases that target platforms actually query.
ASN diversity. A pool heavily weighted toward a single ISP is a detection risk on sophisticated targets. Better pools have IPs distributed across dozens of ISPs per country.
3. Session Control and Targeting Flexibility
A provider without sticky sessions isn’t usable for any workflow requiring authentication or multi-step navigation. A provider without city-level targeting isn’t usable for local SERP tracking or regional pricing research. These are binary requirements, not nice-to-haves — verify them against your actual use case before signing up.
Minimum session controls to require:
- Per-request rotation (default for stateless scraping)
- Sticky sessions with configurable duration (at least 10–30 minute windows)
- Country and city targeting
- ASN targeting (for ISP-specific needs)
- Both HTTP and SOCKS5 protocol support
These should all be exposed through a single endpoint via URL parameter formatting — something likeuser-country-us-city-newyork-session-abc123:pass@gateway:port. If a provider requires different endpoints or separate configurations for each mode, that’s operational friction that compounds at scale.
4. Real Success Rates on Real Targets
Every provider claims high success rates. None of them are measuring success rate the same way, and few are measuring it on the targets that matter to your use case.
A 98% success rate on unprotected HTTP endpoints is meaningless if you’re scraping Amazon, Google, or LinkedIn. What you need to know is the success rate on protected targets — specifically the ones you plan to use.
How to actually test this before paying:
Most reputable providers offer a trial period or small prepaid balance. Use it to run real requests against your actual targets, not “hello world” checks against httpbin.org. Measure:
- HTTP 200 response rate (excluding CAPTCHAs and soft blocks)
- Response accuracy rate (do scraped prices/data match manual verification?)
- Latency distribution — not just average, but P95 and P99
- CAPTCHA encounter rate per 1,000 requests on each target
Run this test at realistic concurrency, not just single-threaded. Provider performance often degrades under concurrent load in ways that single-request tests don’t reveal.
5. Pricing Structure and Traffic Policy
Residential proxy pricing has two variables that matter: cost per GB and what happens to unused traffic.
Cost per GB ranges from around $1 at high volume to $5 for small prepaid balances. The tiering structure varies significantly — some providers discount at $500, others only at $5,000. For teams with predictable monthly volume, the break-even point between tiers is worth calculating explicitly.
Traffic expiration policy is where most providers quietly extract value. A provider that expires your unused traffic monthly means any month where you use less than expected is pure loss. A balance-based model — where you top up and traffic never expires — is structurally better for any workload with variable demand.
At ResidentialProxy.io, pricing starts at $5/GB for small top-ups and drops to $1/GB at $5,000+ — with traffic that never expires regardless of tier. For teams with variable scraping schedules, this matters more than the per-GB rate.
Hidden costs to check:
- Are failed requests billed? (They shouldn’t be — only successful data transfer should count.)
- Are there connection fees or minimum session charges?
- Is there a monthly platform fee on top of usage?
- Do volume discounts apply to a rolling total or reset monthly?
6. Infrastructure Ownership vs. Reselling
A significant share of the residential proxy market is resellers: companies that white-label access to another provider’s network, add a markup, and build their own dashboard on top. This isn’t inherently bad, but it has practical implications.
Support resolution time. When something breaks in the underlying network, a reseller can’t fix it — they can only escalate to the upstream provider. That adds latency to incident resolution. For a production pipeline running continuous data collection, “we’ve escalated to our upstream provider” is not an acceptable support response.
Network control. Resellers can’t tune rotation logic, adjust IP selection algorithms, or add new geographies independently. Feature requests and infrastructure improvements depend on the upstream provider’s roadmap.
Price stability. If the upstream provider changes their wholesale pricing, the reseller’s pricing changes. Resellers often can’t absorb these changes gracefully and pass them to customers abruptly.
How to identify resellers: Ask directly whether they own and operate the proxy infrastructure or whether it’s sourced from a third party. Check whether their engineering team has published technical content about their network architecture. Look at their support team’s ability to answer infrastructure-level questions — resellers often can’t.
7. Support Quality Under Pressure
Support quality is easy to evaluate in the sales process and hard to evaluate when you actually need it. A pre-sales team that responds to trials in minutes may have an entirely different support operation for paying customers, and the difference only becomes apparent during an incident.
What to test before buying:
- Send a technical question through their support channel before signing up. Measure response time and quality of the answer. A 48-hour response with a templated non-answer tells you what post-sale support looks like.
- Ask a question that requires infrastructure knowledge — something like “what’s the ASN distribution of your US residential pool” or “do you support HTTP/2 on the gateway?” Generic answers suggest the support team doesn’t have network-level access.
- Check whether 24/7 support means human coverage or a bot that routes tickets to a queue reviewed during business hours.
For enterprise workloads, a dedicated account manager with technical background is worth paying for. Incident resolution that requires going through a ticket queue when your scraping pipeline is down during a critical data collection window has a real cost.
The Evaluation Checklist
Before paying for any residential proxy provider, confirm:
- Sourcing methodology is documented and opt-in based
- Pool size is sufficient for your target volume with low reuse rates
- City-level targeting is available and geolocation-accurate
- Both rotating and sticky sessions are supported through a single endpoint
- Success rates on your specific targets have been tested with a trial balance
- Traffic doesn’t expire on unused balance
- Failed requests aren’t billed
- Provider owns and operates their own infrastructure
- Pre-sales support response demonstrated real technical knowledge
No provider is perfect on all dimensions. The goal is to know which tradeoffs you’re making before the contract is signed, not after the pipeline is in production.
