Testing methodology

Last methodology review: 5 September 2026 · Next scheduled review: Q4 2026

This document describes exactly how PROXYDECK tests, scores, and ranks proxy providers. We publish it openly so anyone can verify that our process is reproducible, fair, and not influenced by commercial relationships.

1. Independence and funding

The site is funded by affiliate commissions — when a reader signs up to a provider through one of our links, we earn a percentage of the first payment. This has no input into the catalog score. The same testing protocol applies to every provider, partners and non-partners alike.

The Editor's Pick badge is awarded based on test results, not commission size. Providers without an affiliate program can earn it on the same footing — and do.

A place in the catalog is free and is not for sale. A provider can apply through the suggestion form, but applying buys nothing: we test, and a service that fails vetting is not published. Much of the catalog we add on our own initiative, without asking anyone. Paid formats — partner placement, a sponsored review, a landing page — exist separately, are labelled as advertising, and touch neither the score, nor the ranking position, nor the badge.

2. What we test

2.1. Performance metrics

• Success rate — share of HTTP 200 responses across a 1000-request batch against benchmark targets (Amazon listings, Google SERP, a Cloudflare Bot-Fight-protected page).
• Latency — P50 / P95 / P99 round-trip time, measured in parallel from 4 markets (US-East, EU-Frankfurt, UK-London, APAC-Singapore).
• Throughput — concurrent connections per port before success rate degrades.
• Pool diversity — share of unique IPs across 1000 consecutive requests, verified via ASN lookup.

2.2. Geographic coverage

For every provider we sample 50 random IPs per claimed country and cross-check the actual location against three independent sources — IPinfo, MaxMind, and IP2Location. If agreement drops below 80%, the geo is flagged as "mis-declared", which costs points in the geography category. We also log the share of genuinely residential ASNs (not datacenter) for residential providers — a real marker of honest sourcing.

2.3. Pricing transparency

• Landing-page price reconciled against the actual checkout price (sticky-session surcharge, geo-targeting cost, sub-user limits).
• Refund policy and how strictly it's enforced — verified by requesting a refund on a small test payment.
• Whether checkout works without KYC, and the entry price on pay-as-you-go.

2.4. Compliance and use-case tolerance

• Which use-cases are blocked at signup vs. at verification.
• Tolerance for iGaming, affiliate marketing, ad-verification, multi-account work.
• GDPR / CCPA posture — DPA availability, server location, public retention policy.
• Crypto and non-card payment methods, availability of local rails (Wise, SEPA, AliPay).

2.5. Support quality

• First-response time on technical, billing, and refund questions.
• Resolution time for an issue we plant deliberately (typical example: a geo-targeting mismatch in a specific US state).
• Account manager availability and the plan threshold at which one appears.

3. Composite score — 7 categories

Every provider receives a 0–10 score. It's a weighted sum across 7 categories. Weights are open:

Weights sum to 100%. The score-formula caption on every provider page mirrors this table — they're kept in sync.

4. Test environment — 4 markets, weekly cycle

All runs happen on identical infrastructure to remove network-environment bias:

• Hardware: AWS EC2 t3.medium (2 vCPU, 4 GB RAM), identical PHP stack + headless Chromium.
• 4 testing markets: US-East (Virginia), EU-Central (Frankfurt), UK (London), APAC (Singapore). Covers the main timezones and typical latency-divergence patterns.
• Toolchain: in-house benchmark runner (open-source release planned Q4 2026), Chrome DevTools Protocol for JS-fingerprint and Cloudflare-passage checks.
• Schedule: performance — weekly; top-10 deep-dive — monthly; full re-test of the catalog — quarterly.

5. Editor's Pick criteria

The badge goes to providers that meet all of the following:

• Composite score ≥ 9.0.
• Top-3 in their primary category (residential / mobile / datacenter / ISP).
• No known compliance scandals and a clean KYC record over the last 12 months.
• A free trial or pay-as-you-go option with a ≤ $5 entry.
• ≥ 6 months in the catalog — so we have a trend, not a snapshot.

6. Update cadence

• Weekly: performance metrics (success rate, latency, drop-out).
• Monthly: pricing and policy changes, KYC updates.
• Quarterly: full re-test across every composite-score category.
• On-demand: when a provider announces a major change, a scandal surfaces, or ownership shifts.

7. Conflicts of interest and hard disqualifiers

If an editorial team member has a financial interest in a provider beyond a standard affiliate commission (equity, paid consulting), they recuse themselves from that provider's evaluation. We disclose every such case publicly.

Hard disqualifiers — providers that do not enter the catalog regardless of technical performance:

• Legal ties to the Russian Federation (founders, registered office, infrastructure, accepting Russian payment rails as a primary method).
• Documented botnet origins of the IP pool (911.re, RSocks, Faceless, and related ecosystems).
• Demonstrated IP-farming via malware or deceptive software.

8. Corrections and disputes

If you spot a factual error or want to dispute a score:

• The "Report an error" button on any review page.
• Email [email protected] with evidence (screenshots, raw logs, links to official announcements).

We respond within 48 hours. Corrections are published within 5 business days when the evidence holds.

9. How a provider enters the catalog

Providers can request inclusion by emailing [email protected]. We do not accept payment for inclusion or score adjustment. Sponsored formats are clearly labelled — see the About page.

10. What we don't do

Transparency outweighs claims, so:

• We don't publish scores without a live test — providers that haven't cleared the full cycle appear with a verified=0 marker.
• We don't use synthetic benchmarks — every run hits real public targets.
• We don't accept payment for rewriting the Pros / Cons.
• We don't leave scores stale: if a provider hasn't been tested for over 90 days, a "stale data" warning appears on its page.

If you have a hypothesis about a test or metric worth adding, the "Suggest a new pattern" button below goes straight to our editorial queue.

Got your own protocol — or spotted a gap in ours?

Tell us — propose new test patterns, metrics, edge-cases. Anything that holds up goes into the next quarterly re-test. We also publish the ideas we discard, with reasons.