Choosing a Provider

Open ten provider homepages and the claims blur together: tens of millions of IPs, nearly every country, high success rates. None of those lines says how a network behaves against one buyer's targets, in that buyer's locations, at that buyer's volume. A scorecard and a one-week pilot answer that question.

A 12-point scorecard

The scorecard turns a vague impression into a number. Score each row from 1 (poor) to 5 (excellent), then double the rows that matter most for the job. A scraping team might double coverage and pricing fit. A team managing logins might double session control and proxy types.

Criterion	What to check	A good sign
1. Pool and coverage	IPs in the countries and cities the job needs	Per-country figures, not one global total
2. Proxy types	Which of the four types are sold	The type needed now, plus the next rung up
3. Targeting	Country, region, city and carrier options	The needed level is included in the plan price
4. Session control	Sticky duration and rotation settings	Session length can be set per connection
5. Pricing fit	Billing model against the shape of the traffic	Low cost per successful request in a pilot
6. Trial and refund	How to test before committing	Trial or refund terms stated up front
7. Sourcing transparency	Where residential and mobile IPs come from	A public page on consent and opt-out
8. Documentation	Setup guides and code samples	Public docs, readable without an account
9. Support	Channels, hours and speed of reply	A technical answer within hours during the pilot
10. Dashboard and API	Usage stats, sub-users and automation	Usage broken down by day and by target
11. Compliance and KYC	Acceptable-use policy and customer checks	Clear rules and a real verification step
12. Reputation	Track record and independent feedback	A named company with years of operation

Rows 1 to 6 can be scored from a pricing page and a trial. Rows 7 to 12 take reading and a few emails. Skipping the second half is how teams end up with a cheap plan from a provider that disappears mid-project. The appendix has a compact worksheet to copy for each provider on the shortlist.

Pool size, handled with care

Pool size is the number buyers quote most and understand least. Every figure is vendor-stated, and no outside body audits it. Three problems follow.

First, vendors count differently. One counts every IP seen over a month. Another counts IPs online at a given moment. Residential devices come and go, so the two methods give very different totals for the same network.

Second, networks overlap. Resellers draw on the same upstream pools, and one household device can sit in several networks at once. Adding up the claims of ten vendors does not give ten separate pools.

Third, a global total says nothing about one place. A job that needs IPs in Lisbon is served by the IPs in Lisbon. A hundred million addresses elsewhere do not help.

Success on the real targetMeasured in a pilot

The share of requests that return the right content from the sites the job needs.

Coverage where the job runsPer country and city

Enough IPs in the needed locations that the same few are not reused all day.

Session stabilitySticky means sticky

Sessions that last as long as configured, without dropping mid-task.

Headline pool sizeA rough signal of scale

Useful for sorting very small networks from large ones, and little else.

How clean the IPs are matters more than how many exist. An address that other customers hammered against the same retail site yesterday arrives already flagged. No vendor publishes that figure, which is one more reason to run a pilot.

Ethical sourcing and KYC

Chapter 2 explained that residential and mobile pools are built from real people's devices, and why their consent matters. For a buyer comparing vendors the question becomes practical: how to score rows 7 and 11 from the outside.

Row 7 is scored on what a provider publishes. The useful evidence is a public page. It should say how devices join the network, what the owner sees before agreeing, and how to leave.

Row 11 looks in the other direction, at how the provider screens its own customers. Customer identity checks, known as KYC, can mean a company email, an ID check or a short call. Some providers also block sensitive targets, such as banking or government sites, until a use case is approved.

The checks slow signup down, and buyers often resent them. They are still a good sign. A provider with no checks sells the same pool to anyone, and the abuse that follows lands on IPs every other customer shares. A provider that asks what the traffic is for is policing its own network.

Ten example networks

The table lists ten networks as examples drawn from the ProxyLook directory. It is in alphabetical order and it is not a ranking. Pool figures are vendor-stated and are reproduced as published.

Network	Proxy types	Vendor-stated pool
Bright Data	Residential, datacenter, ISP, mobile	150M+ IPs
Decodo (formerly Smartproxy)	Residential, datacenter, ISP, mobile	125M+ IPs
IPRoyal	Residential, datacenter, ISP, mobile	32M+ IPs
NetNut	Residential, datacenter, mobile	85M+ residential, 5M+ mobile
NodeMaven	Residential, ISP, mobile	30M+ residential, 250K+ mobile
Oxylabs	Residential, datacenter, ISP, mobile	177M+ IPs
ProxyEmpire	Residential, datacenter, ISP, mobile	30M+ residential, 4M+ mobile
Proxy-Seller	Residential, datacenter (IPv4 and IPv6), ISP, mobile	20M+ residential
SOAX	Residential, ISP, mobile	155M+ IPs
Webshare	Residential, datacenter	80M+ residential, 30M+ datacenter

The pool column makes the earlier point by itself. Some vendors publish one total and others split residential from mobile, so the figures do not line up. The types column is more useful. Six of the ten list all four proxy types, which lets a team climb the ladder without changing vendor.

The full list of more than 200 providers is at proxylook.com/all-providers. ProxyLook publishes how it scores them on its methodology page, and the compare page puts candidates side by side.

A seven-day pilot

A pilot is a small, timed experiment with a pass mark agreed in advance. One working week is enough to separate two or three candidates.

1. Day 1: write the brief. List targets, locations, daily volume and session needs. Set a pass mark for success rate and for cost per 1,000 good pages.
2. Day 2: shortlist and sign up. Use the scorecard to cut the list to two or three. Open trials and complete any verification.
3. Day 3: smoke test. Confirm authentication, exit IP and exit location for every country the job needs.
4. Day 4: run at low volume. Send a few hundred requests per target with the same script for each provider. Log success, response time and traffic used.
5. Day 5: run at production shape. Use full concurrency at the hours the job will run. Watch for success falling as volume rises.
6. Day 6: test sessions and support. Hold sticky sessions for as long as the job requires. Send support one real technical question and time the reply.
7. Day 7: score and decide. Compute cost per 1,000 successful requests, fill in the scorecard and file the raw logs.

The raw logs matter later. When success drops six months on, they show what normal looked like.

Red flags

One of these signs alone is a reason to ask questions. Several together are a reason to walk away.

• No explanation of where residential or mobile IPs come from.
• A pool figure with no country breakdown and no definition of what is counted.
• No trial, no refund and a large minimum purchase.
• "Unlimited" plans with no published fair-use terms.
• Marketing that promises accounts or scrapers will never be detected or banned.
• No company name, address or terms of service on the site.
• Documentation hidden until after payment.
• Support that answers sales questions quickly and technical questions never.

Key takeaways

• Score providers on 12 criteria and weight the rows that matter for the job.
• Pool sizes are vendor-stated, counted in different ways, and overlap between networks.
• Success on the real target and coverage in the needed locations outweigh headline pool size.
• Consent-based sourcing and customer verification protect the buyer as well as device owners.
• A seven-day pilot with a pass mark agreed in advance beats any homepage claim.
• The ten networks named here are examples, not a ranking, and the full list is at /all-providers.