How to Test a Proxy (2026): Speed, Location and Success Rate | Coupo9
How-to8 min read
How to Test a Proxy Before You Buy More Traffic
A short test you can run yourself: confirm the exit IP and location, measure speed and success rate against your own target with curl or Python, then work out your real cost per successful request before you buy more traffic.
Knowing how to test a proxy saves more money than any coupon on a proxy plan. A provider's homepage quotes its own success rate and speed, measured on its own targets. Your scraper, your login flow or your price checker hits different sites, and those numbers can look very different there.
The good news is that a useful test takes under an hour and a few hundred megabytes. You check that the exit IP is where you asked for it, time a batch of requests against your real target, count how many come back clean, and turn that into a cost per successful request.
Do it inside a provider's trial or refund window and a bad fit costs you little. This guide walks through the method step by step, so by the end you'll know exactly how to test a proxy before the next top-up.
Check the exit IP and country first. A wrong location ruins every other number.
Time 100 to 200 requests against your own target, not a speed-test site.
Count success as a real page, not just an HTTP 200.
Divide what you paid by successful requests to get your true cost.
Run the test inside the trial or refund window, which can be as short as 2 days.
What you're actually measuring
A proxy test answers four questions, in this order:
Location: does the exit IP sit in the country or city you asked for?
Speed: how long does a request through the proxy take, end to end?
Success rate: what share of requests return the page you wanted?
Cost: what does each successful request cost once failures are counted?
Most buyers stop at question two, because speed is easy to see. Success rate matters more. On a per-GB plan you pay for blocked pages, captchas and timeouts that still move data, which our guide to how proxy pricing works covers in detail.
Pick your test target before you start. Use the site you actually plan to work with, or the closest public page on it. A proxy that loads a generic speed-test site quickly can still fail half the time on a retailer with strong bot protection.
Step 1: Confirm the exit IP and location
Start with a request that tells you what the website sees. An IP lookup service such as ipinfo.io returns the IP, city, country and network owner as JSON.
With curl, the -x (or --proxy) option sends a request through a proxy:
Run it five times. With a rotating residential proxy you should see five different IPs, all in your target country. With a sticky session, the IP should stay the same until the session ends. Our guide to rotating vs sticky proxies explains how to set the session string for each.
Check the org field too. If you paid for residential IPs and the lookup shows a hosting company's network, you're not getting what you bought. Datacenter and ISP proxies will show a business network, which is expected. If you're unsure which type you need, our guide to proxy types compares them.
Step 2: Run a proxy speed test with curl
curl's -w (write-out) option prints timing and status after each request. Three variables do most of the work: http_code for the status, time_starttransfer for the time to the first byte, and time_total for the whole request.
Wrap that line in a shell loop of 100 runs and save the output to a file. You now have a status and two timings for every request.
Don't judge on the average alone. Sort the time_total column and look at the middle value and the slowest tenth. A proxy with a fine average and a long tail of 20-second requests will stall a scraper that waits on each page.
Run the same loop once without the proxy, from your own connection. That gives you a baseline, so you can see how much time the proxy adds rather than how slow the target is.
Step 3: Measure success rate with Python
A 200 status isn't proof of success. Many sites return a 200 with a captcha or a "please verify you're human" page. For a proper count, check the content.
The Python requests library takes a proxies argument on any request. A minimal test looks like this:
import requests, time
proxies = {"http": PROXY_URL, "https": PROXY_URL}
r = requests.get(TARGET_URL, proxies=proxies, timeout=30)
Mark the run a success only if r.status_code == 200 and a word you expect on the real page, such as the product name, appears in r.text.
Record the time, the status and len(r.content), then loop 100 to 200 times.
Always set a timeout. Without one, a single dead connection can hang the whole test.
The byte count matters for cost. Add up len(r.content) across the run and you have a rough idea of how much traffic a page uses. It undercounts, because headers, redirects and page assets you didn't download aren't in it, so check the provider's dashboard usage after the run as well.
Step 4: Work out cost per successful request
This is where the test pays for itself. The formula is simple:
Cost per successful request = traffic cost of the test ÷ number of successful requests
Here's a worked example with clearly hypothetical numbers. Say a test of 200 requests used 300 MB on a plan that costs $3.50 per GB. That comes to about $1.05 of traffic, counting 1 GB as 1,000 MB. If 150 requests succeeded, each good page cost you about 0.7 cents. If only 90 succeeded on another provider at a lower per-GB price, the cheaper plan can still cost more per page you can use.
Compare providers on this number, not on the price per GB. Then multiply by the pages you need each month to estimate a real budget. Our list of cheap and reliable proxy providers makes the same point: the cheapest proxy is the one that works on your target.
Use the trial or refund window
Every provider sets its own limits, and some are tight. Here's what each official page said when we checked on 13 November 2026.
Provider
How you can test
Limits that matter
Decodo
Free 3-day trial with 100 MB on pay-per-GB plans
Card needed; the plan starts after the trial unless you cancel. Using the trial rules out the 14-day money-back
Decodo
14-day money-back on a first self-service purchase
Not more than 20% of the plan's traffic or 1 GB, whichever comes first
Webshare
Free plan: 10 shared proxies, 1 GB a month, no card
Datacenter only, so it won't tell you how residential IPs perform
Webshare
Refund within 2 days of payment
Under 0.1 GB used on rotating residential, under 1 GB on proxy server or static residential
NodeMaven
Paid trial: $3.50 for 750 MB
No free trial
IPRoyal
Refund for service problems support can't fix
Request within 24 hours of purchase; not available after more than 100 MB is used
The pattern is clear: plan your test before you pay. With Webshare's rotating residential refund limit of 0.1 GB, a test of a few hundred heavy pages could already cost you the refund. Keep the test small, or pick a trial with enough data for your target.
Decodo's help centre adds one note for Indian cards: the automatic charge at the end of a trial can't go through without your confirmation, so you buy the plan from the dashboard once the trial ends.
A one-hour test plan
Put the steps together and a full test fits in an hour:
Write down your target URL, the country you need and a word that only appears on the real page.
Start the trial or buy the smallest plan, and note the time so you know when the refund window closes.
Run five IP lookups and confirm the location and network type.
Run 100 requests without the proxy for a baseline.
Run 100 to 200 requests through the proxy with the Python check.
Note the success rate, the middle and slowest response times, and the dashboard traffic used.
Work out cost per successful request and compare it with any other provider you tested.
Decide before the window closes: scale up, switch provider or ask for the refund.
Test at the time of day you'll actually run the job. Residential pools behave differently at peak hours.
Our take
The single most useful habit when you learn how to test a proxy is measuring on your own target, not trusting the provider's average. A 99% success rate on a homepage tells you nothing about the site you care about, and a cheap per-GB price means little if half your requests come back as captchas.
Keep the test small enough to stay inside the refund limits, count success by content rather than status code, and compare providers on cost per successful request. That's the same pay-only-for-what-works rule our guide to saving money on software subscriptions applies to every tool you rent.
If you want a free starting point, Webshare's free plan is enough to practise the method. For residential tests, Decodo's 3-day trial and NodeMaven's $3.50 trial give you real data to work with. Check current offers on Decodo coupons, Webshare coupons, NodeMaven coupons and IPRoyal coupons, or browse every provider in our proxy deals hub. One hour of testing is the cheapest part of how to test a proxy properly, and it can save you a month of wasted traffic.
Trial and refund terms checked on 13 November 2026 from each provider's official pages: [Decodo trials](https://help.decodo.com/docs/trials), [Decodo refund policy](https://decodo.com/faq/billing/can-i-get-a-refund), [Webshare pricing](https://www.webshare.io/pricing), [Webshare refunds](https://help.webshare.io/en/articles/8370527-requesting-a-refund), [NodeMaven pricing](https://nodemaven.com/pricing/) and [IPRoyal residential refund policy](https://help.iproyal.com/en/articles/7215815-what-is-your-refund-policy-for-residential-proxies). Command options from the [curl manual](https://curl.se/docs/manpage.html) and the [Requests proxies documentation](https://requests.readthedocs.io/en/latest/user/advanced/#proxies). Terms change, so check the live pages before you buy.
Topics
proxies
C
Written by
Coupo9 Editorial
Every statistic in our guides names its source, and prices come from the brand's own pricing page. When the facts behind a guide change, we update it or take it down.
How Proxy Pricing Works: Per GB, Per IP and Per Port
Proxy pricing comes in three units, and each one punishes a different mistake. Here's how per-GB, per-IP and per-port billing work, how to estimate your traffic before you buy, and which providers let unused gigabytes expire, with eight providers' prices checked on 11 Nov 2026.