---
title: "How we test"
url: https://proxy.wiki/how-we-test/
type: Page
author: ""
published: 2026-08-19
updated: 2026-08-19
site: proxy.wiki
license: CC BY 4.0 — quote freely with attribution to https://proxy.wiki/
---

# How we test

> The rubric, the data statuses, and what our numbers can and cannot tell you.

Every figure labelled _measured_ on this site links back to this page. If a claim cannot be traced to a method described here, it is not a measurement and we do not present it as one.

## The three data statuses

| Status | What it means | What may appear |
| --- | --- | --- |
| Measured | We ran the test ourselves | Performance figures, plus the date, the sample size and a link to the method. All three are required together. |
| Vendor-reported | The provider published it | Vendor figures, each labelled as a claim, each linking to the page that makes it. |
| Not yet tested | We have not run this provider | No performance figures at all. The page says so in a visible box rather than quietly omitting the section. |

## The scoring rubric

Reviews carry an overall score out of 10 and up to five sub-scores. A sub-score we have not assessed is omitted, not defaulted to zero or to the average.

| Dimension | What it covers |
| --- | --- |
| Performance | Success rate and response time against real targets, where we have measured them. |
| Pool | Breadth of geography and targeting granularity — country, city, ASN. |
| Features | Protocols, rotation control, session stickiness, concurrency limits, API and dashboard. |
| Value | Entry price against what you actually get, including minimum spend and billing model. |
| Support | Responsiveness and usefulness of support during our own account setup. |

## What a benchmark can and cannot tell you

This section matters more than the numbers themselves.

- **One request is one sample.** Our proxy checker shows a single exit node at a single moment. It is a diagnostic, not a benchmark, and we label it that way on the tool itself.

- **A rate needs a denominator.** “98% success” is meaningless without knowing whether that is 100 requests or 100,000, and against what. We publish the sample size beside every rate or we publish neither.

- **Results are target-specific.** A pool that sails through one site may fail completely against another with different anti-bot tooling. We name the targets we tested.

- **Results decay.** Pools change, routes change, prices change. Every figure carries the date we recorded it, and reviews untouched for six months are flagged internally for re-checking.

- **We do not count IPs.** Nobody outside a provider can verify a pool size. We report the claim, link to it, and label it as a claim.

## Conflicts of interest

Where a commercial relationship exists with a provider, the disclosure block appears on that page. Commission has no effect on a score or a ranking. See [how we make money](/disclosure/) and our [editorial policy](/editorial-policy/).

## Tell us we are wrong

If you can show a figure here is inaccurate, we want the correction. Send the evidence and we will re-check, update the page, and record what changed.