Node HandbookNODE HANDBOOK中文
Providers · evaluation

Evaluating a proxy service: six checks instead of a ranking

A ranking answers “which is best” without knowing what you need. Six checks answer “does this suit my traffic, at my hour, on my network”, which is the only question with a usable answer.

Node Handbook editorial team ·

The six checks

CheckWhat to establishWhere to look
1. Billing ruleQuota, multiplier scope, reset boundary, what happens at the limitThe written terms, not a chat reply
2. Exit verificationObserved exit addresses, and whether your target services accept themYour own request plus two IP data sources
3. Peak behaviourSustained throughput at the hour you use it, with a no-proxy controlYour own measurement on a weekday evening
4. TermsRefunds, renewal, fair use, and whether rules can change mid-termThe published terms, saved at purchase
5. SupportWhether a question gets a substantive answer, and how fastAsk one real question before paying
6. Payment riskHow much a failure would cost you, given the term you are buyingYour own decision, made before purchase

Why this order

Billing comes first because it determines whether the plan can do your job at all. A generous but mispriced quota — a high multiplier on the route you need — makes every later measurement irrelevant. Exit verification comes second because a plan whose exits your services refuse has no value regardless of speed.

Peak behaviour third, because it is the most common disappointment: everything works in the trial, and the first busy evening is the real test. Terms fourth, because they determine what happens when any of the first three fails.

Support fifth, because it is cheap to test and expensive to discover late. Payment risk last, because it is the conclusion, not an input.

A one-week evaluation

  1. Day 1: buy the shortest term. Record quota, multiplier scope and reset boundary. Import one profile and verify one real request.
  2. Day 2: read the observed exit from two sources; test the services you actually use.
  3. Day 3: send one support question and record the response time and whether the answer addresses the question.
  4. Day 4–5: measure peak-hour throughput on a weekday evening, three runs, with a no-proxy control.
  5. Day 6: read the terms as published and note the refund and renewal clauses.
  6. Day 7: decide. If any of the first three checks failed, do not extend the term.

Signals that deserve weight

  • Terms that are written, dated and internally consistent.
  • Announcements that are archived rather than edited after the fact.
  • Support answers that name a cause and a next step.
  • Published measurements that state their conditions, including the hour and the target.

Signals that deserve less weight: unqualified superlatives, “unlimited” without a policy, review lists with affiliate links, and speed claims with no timestamp.

Signals that should stop the purchase

  • A requirement to install a root certificate to use the service.
  • Payment instructions that arrive only by direct message, or a different payment address because “the site is down”.
  • No written terms at all.
  • Pressure to buy a long term before any trial.

What to record

  • The six checks, each with a date and a one-line result.
  • The observed exit addresses and the sources used.
  • Peak-hour numbers, all runs, plus the control.
  • Terms as saved at purchase.
  • The maximum you were willing to lose.

Scope

This page provides an evaluation method. It names no provider, ranks nothing, and publishes no test results for any service. Its only claim is structural: these six checks cover the decisions that a ranking cannot make for you.