Methodology: how we choose picks, and what we don’t do

Why there is no overall ranking

An overall ranking needs weights: how much is price worth against targeting, or a recent test against a refund window? Change the weights and the order changes. A buyer who needs city targeting and one who needs a week-long session aren’t helped by the same number, and a single score hides exactly the trade-off each of them needs to see.

So we don’t publish scores, weighted totals or ordinal positions. We name a pick for each use case where the evidence produces a clear winner and give a runner-up where one is useful. The section numbers on the picks page are for navigation.

How a pick is chosen

1. Start from buyers’ needs, not providers. Use cases come from the questions buyers actually face: price at different volumes, billing terms, targeting depth, session length, measured performance, pool size, customer vetting.
2. Write the rule down first. Each use case gets a one-sentence decision rule, published below.
3. Apply it to confirmed facts. Prices and terms from vendors’ pages, checked on 17 October 2026; performance from dated third-party tests.
4. Pick whoever the facts put on top, and name a runner-up where a different trade-off would suit some readers.
5. No winner, no pick. If the evidence doesn’t separate providers, we don’t invent a use case to give one of them a win. Four of the eleven providers currently have no pick.

Every use case: rule and evidence

The same nine use cases as the picks page, in the same order. Figures are as of October 2026; sources are cited on the picks page and in each provider’s review.

Where the facts come from

Vendor pages

Prices, plans, minimums, expiry, refund terms, targeting, protocols and session limits come from each provider’s own pricing pages, documentation and terms, checked on 17 October 2026. Every review links the vendor’s pricing page, and prices are labelled “as of October 2026”.

Proxyway

Performance figures — success rates, response times, unique-IP counts — come from Proxyway’s provider reviews and its Proxy Market Research 2026¹. Most residential tests cited here ran in April 2026; Evomi’s Core pool was tested in July 2026²; Bright Data’s latest published residential test dates from March 2024³. We also cite Proxyway for some facts vendors don’t publish, such as KYC requirements and support hours, and say so where we do.

Synthient

Independent ratings of proxy networks’ consent practices. We found a rating for PacketStream only⁴.

News

Proxyway’s news report on the Decodo rename⁵.

Trustpilot

Used only for patterns of complaint and for rating suspensions — never as a score — and only where it adds something: Proxy-Cheap and IPRoyal^6,7.

HTTP Archive

Page-weight medians for the “what a gigabyte buys” arithmetic⁸.

Calculations — cost at a given volume, cost per 1,000 pages, retry multipliers — are our own arithmetic on those figures, and each table says how it was worked out.

Reading the evidence

• Test dates differ. A March 2024 result says little about a network in October 2026. We print the test date with every figure.
• Infrastructure tests aren’t target tests. Most figures come from requests to a small test page through each provider’s gateway. They measure speed and stability, not success against a particular anti-bot system.
• Unique-IP counts are samples, gathered over fixed periods and request volumes, not full pool sizes.
• Vendor claims are claims. Pool sizes, “ethically sourced” labels and similar statements are reported as the vendor’s words.

What we couldn’t confirm

We leave out, or label, anything we couldn’t verify. As of this update that includes:

• Bright Data’s sticky-session maximum and traffic expiry.
• SOAX’s sticky-session maximum on its own pages (Proxyway reports up to a week).
• Whether Webshare’s unused traffic rolls over, and its current yearly prices.
• Proxy-Cheap’s discount terms, targeting options, expiry and session limits.
• Infatica’s pay-as-you-go minimum; Rayobyte’s expiry terms on its own pricing page.
• IPRoyal’s prices between 10 GB and 10 TB.

What we don’t do

• We don’t publish scores, star ratings, rankings or “overall winner” badges.
• We don’t publish user testimonials, invented bylines or credentials.
• We don’t add rating markup to search results; there are no user reviews on this site to rate.

Updates

Prices change often. Before each update we re-check every vendor page, re-apply each rule and change the “prices verified” date shown on every page. If a provider’s terms change between updates, the affected pages are updated when we learn of it.