You’d think that testing hosting services is a pretty straightforward process. That it’s a simple checklist of features, coupled with standard performance tests that just can’t lie. But that’s not how things are. At all. Most reviews are not as straightforward as what you’ll find in the most transparent web host comparison.
Because people are involved in every step of the way, not only can these tests lie – they often do. It’s both expensive and demanding to actually sort through all the options a web host offers, test it thoroughly, and come up with helpful information that’s relevant to you – the user.
And that’s why many “testing” websites use a shortcut. An easy way out; one that’s almost invisible. They work together with the web hosts, nurturing a mutually beneficial relationship.
“Wait, what?” you must be thinking. “If they work together with the hosts, how can they be trusted?” Well, they can’t. Welcome to the reality of web hosting server performance tests.
Read on to find out exactly who has been fooling you and how. After that, I’ll explain the correct way to do things, and how it can help you choose the best web host for you.
The Problem with the Reviewer-Hosting Relationship
You see, any website that reviews hosting services (actually, reviews anything) makes money from doing so. Hosting companies are always searching for new clients, and they pay an affiliate commission for every new client that clicks a link in a review.
This is a pretty good idea at its core, one where everybody should benefit: the reviewer gets rewarded for finding an awesome service, the user enjoys it, and the company grows and gets better.
Small problem. It doesn’t take human nature into account.
What many people discovered is that it’s much more profitable to just hand-pick the web hosts that provide the best affiliate deals, and write a rave review detailing the endless joy these hosts brought into their life. The amazing loading speeds their site visitors enjoy. The fact that it’s never had any downtime. And they’ve been testing!
Well, have they been testing? Usually, not at all. It’s too easy to get super-technical with the terminology, make up some numbers (or just copy them from the web host’s homepage), and wrap it up in a convincing way. If you’ve ever seen the notorious Bluehost on a “best services” list, you can be positively sure that it paid its way into it.
The people behind “testing” websites like this are basically scammers. They lie to your face and run away with the money. The only good news is that while they make up the majority of online hosting reviews, they’re easy to spot once you realize that they’re not backed by any data.
But even data can lie. Which brings us to the second type of performance test offenders.
Why Sometimes Even Data Can’t Be Trusted
Due to the rising demand for honest, unbiased web hosting performance tests, a new type of testing website arose. One that collects every bit of possible data, runs a dozen different types of tests, and really monitors each web host’s performance for months.
Sounds amazing, right? It really is. We do this too, and are very proud of our work. But some other websites were so hungry for data, they signed a deal with the devil.
Let me explain.
After setting up a hosting plan with one of the vendors, installing website software (usually WordPress), and connecting to a domain, the site becomes accessible online. In evaluating an online website’s performance, we’d like to know many things:
How fast does it load? How fast does it load from different parts of the world? How powerful is the plan, and how many resources does the website have available? What is the website’s uptime?
And… How does the website perform under high loads, when many visitors are requesting it at the same time?
Now, this last question is where the overreaching begins, and where the testing websites break your trust. Not because it’s not an important question (on the contrary, it definitely is), but because testing it is much more problematic.
This type of performance test (usually called a stress test or a load test) is done by sending multiple requests to the server asking for the hosted website. It’s meant to simulate situations where multiple users (hundreds, thousands, even more…) are trying to access your website at once, and to give you an idea of how your website performs under pressure.
But trying to run a stress test on a shared hosting plan doesn’t work. Ever heard of a DDoS (distributed denial-of-service) attack? It’s an attack where multiple requests are made to a server in an attempt to overload it and bring it offline. Quite similar to a stress test… only a tad more malevolent.
To protect their users from such attacks, almost all web hosts today have powerful DDoS defenses in place. If the server senses that an unusually large number of requests is coming through, it’ll simply block and deny them, shutting down the attempt.
So how can these testing websites stress-test the hosts without triggering the defense? They can’t. They have to contact the hosts and ask for all security measures to be removed. But by doing this, they reveal who they are to the hosts, and from this point on, the data they collect is meaningless.
Once you’ve explained to the web host what you’re trying to do, you might as well go home. There’s no validity to your results anymore. The web host knows what’s going on, and it has an incentive to play around with the results.
Even if it doesn’t manipulate the server setup, the host can’t let anybody run stress tests on a populated shared server – if the server crashes, thousands of paying users can suffer downtime. The only way for a host to let you run a load test without risking the other shared accounts would be to give you an empty server.
Are results drawn from testing an empty server relevant to your actual user experience? Nope. Can you trust any data that was produced by revealing that you’re a tester? Definitely not.
Every time you see a beautiful graph or an informative table detailing a shared plan’s performance under a stress test, you can be sure the host was involved in the process. That the tester and the testee were in cahoots. That’s reason enough to ignore this data completely.
Now, these testing websites say that they have “good faith” in the hosts… that they trust them to play fair. Seriously? How can you have good faith in anybody shelling out 200% affiliate deals for their products? Do not fall for these reviews. Learn to spot the warning signs and don’t get fooled!
The Correct Way to Do It
Was I harsh up until now? Maybe. A teeny bit. But honestly, folks, there’s a good reason for that. I’ve been in the industry for a while, and I’ve seen all kinds of ”tests” and “checks” and promises and whatnot. It looks convincing, but in the end, it’s just more of the same. And it’s your website that ends up being slow, unreliable, and unresponsive.
That’s why here at Website Planet, we do things differently. We never, ever, ever, reveal who we are to the web hosts. We always maintain complete secrecy, posing as regular Joes and Janes just trying to get a website up. This is the only way to gather accurate data that is relevant to real users.
Our testing process isn’t complicated. It doesn’t have to be. We simply test everything we can test, striving to emulate the experience a normal user would have. We also test more than just performance. We test the support – the actual support experience, the features that are provided, how easy to use is the service, and what value you get for your money.
Regarding the server performance tests, this is how we do it.
We sign up for the basic shared hosting plan the web host offers. In the few cases where the basic plan can’t handle a WordPress website, we sign up for the first plan that can.
Using the hosting control panel and any automatic installation tools provided by the host, we install WordPress. We then upload our customized landing page, The Autonomous Shoes.
What are autonomous shoes? Who knows. But it’s a complete landing page that we’ve designed to mirror a real website as closely as possible. It has HD images, cool scrolling effects, some blog posts, and a lot of text content.
Many other testing websites only test a bare WordPress theme, with no images or actual functionality. This is silly, because your website isn’t going to be anything like that. The results we get actually give you a good idea of how your real website will perform.
After we get our website online, we contact support with a simple question – “My website is slow. Can you help me optimize it?” The reason we do this is to give the web host a chance to show their expertise, and to point us at any handy tool that can help improve speeds. This can be a caching plugin, a CDN, an image optimizer, or whatnot.
We follow the web host’s advice, as long as it’s simple to implement. Our thinking is that we’re not going to do anything that a new user wouldn’t be able to do.
Now that our website is live and optimized to some degree, we start running performance tests. These include the UptimeRobot uptime monitor, a GTmetrix speed loading and optimization test, the Sucuri Load Time Tester tool, and local tests performed on are own testing servers.
Let’s go into greater detail.
Uptime is the percentage of time where the website is available to visitors online. Everybody is aiming for 100% uptime, but that’s almost impossible to achieve. Web servers, like computers and mobile phones, are always prone to buggy behavior. These minor glitches can result in the server going offline for a short period of time, hopefully no longer than a couple of minutes.
A lot of web hosts guarantee 99.9% uptime, backing this up with a promise to refund you if they can’t make that figure. We appreciate these guarantees, but feel like it’s better to test everything ourselves. We’ve had some sad experiences with hosts that were as reliable as a drunk driver.
UptimeRobot, the monitor that we use, sends the website we’re testing an HTTP request every minute. We keep our websites up for months and provide you the raw data and the screenshots taken from these monitors. Very few of the hosts we test score more than 99.99%, but some do. And we sure appreciate it.
GTmetrix has test servers located worldwide: Dallas, US; Vancouver, Canada; London, UK; Hong Kong, China; Mumbai, India; São Paulo, Brazil; and Sydney, Australia. We use these servers to request our test website, load it, and generate a detailed breakdown of loading time and present optimizations.
We always run GTmetrix tests on our website over a period of multiple days to cancel out any flukes or irregular behaviors. We then aggregate the results and draw insights regarding the web host’s performance.
The location of the test server used depends on the location of our web host’s data center. We choose the closest test server possible, understanding that it gives you the best idea of how well your website can perform.
Through these tests, we can see which hosts provide the best loading speeds, and which ones have built-in caching, compression, and other optimization tools. The final score doesn’t lie.
The Sucuri Load Time Tester
The Sucuri Load Time Tester utilizes a global network of testing servers to analyze a website’s performance worldwide. The tool logs the time in seconds it took for the first connection to be made, for the first byte to be sent, and for the total loading time.
It does this separately for each location, and at the end also provides a global performance grade. By using it we gain valuable insights as to how a website will perform for various global audiences.
Our Own Testing Servers
The tools I’ve introduced are all incredibly useful, but they don’t have active locations in every country around the globe. Luckily, we do. Our global web host comparison project is conducted by local experts who in addition to using all aforementioned tools, also employ the use of their own test servers.
The figures we get from these tests are the most interesting we have, and you just can’t find them anywhere else.
Our local experts use a powerful server located in their home country and connected to a high-speed internet connection, to test actual loading times on Google Chrome. They use the DevTools supplied with the browser to clock response times and the total time it takes for the website to load. The data accumulated this way is the most precise measurement of how a specific web host will perform in a specific country.
Why Country-Specific Rankings Are the Only Thing That Matters
If you ask someone “What’s the best host?” and they give you an answer, any answer, it’s meaningless. Sadly, and despite the location of the target audience being the most important variable in deciding what web host is best, most “testing” websites just give a general, one-size-fits-all answer.
We don’t. When we’re testing a web host, we test it in a specific country. We upload content in that country’s language, and use a local expert to interact with the technical staff. We gather all available data, combining features, ease of use, performance, support, and pricing, to come up with the most conclusive scores.
Is it expensive? It sure is. For example, when we wanted to test the popular Hostinger, we had to sign up to the service over two dozen times, each time for a different country. We checked how the host performed in Denmark, in Canada, in Japan, in Poland, and in many other places.
Each time, the results we got were a bit different. For example, in Israel, where local companies leave quite a bit to be desired and where local servers are rare to find, Hostinger made it to the #1 spot, providing a great combination of loading speeds and features, and at super-competitive prices. Then again, in Russia, where Russian hosts maintain data centers much closer to their target audience, Hostinger only came in at #4.
We never know who’s going to win a web host comparison, or which host is going to come out on top in a certain country. We have no incentive to play around with the results, and we never intend to do so. We simply look at the data, listen to our experts, and provide you with the best possible information.
Check out our main hosting comparison page for the current top-performing hosts in the US, or venture into any of the localized versions for country-specific rankings.