My recommendations are based on a carefully thought-out methodology that includes ongoing collection of performance data from a collection of test websites with many of the most popular shared web hosts. I share performance data via my web hosting performance dashboard and my post discussing the results of LoadImpact tests.
I try to be extremely transparent about my evaluation process. In-the-weeds details about my methodology and its limitations can be found below the recommendations. I have a financial relationship with some web hosts. Specific details about the relationships can be viewed by expanding the “Relationship disclosure” sections on this page.
Great prices without gimmicks
In an industry where gimmicks are extremely common, Namecheap stands out for its straightforwardness and transparency. Customers still get Namecheap’s ultra-low prices when purchasing on a monthly billing cycle without any long-term commitment.3 I have not seen Namecheap engage in the aggressive (and sometimes deceptive) upselling that is common in the web hosting industry.
Individuals that are willing to pay for top-tier performance may be better served by other web hosts. That said, I think Namecheap’s performance will be more than adequate for the majority of websites.4 Read more of my thoughts on Namecheap…
SiteGround’s services are on the expensive side. If you have a tight budget or don’t need top-notch performance, there are cheaper options that can still deliver good performance for websites with modest needs. Read more of my thoughts on SiteGround…
Hosts to be cautious of
Endurance International Group brands
The Endurance International Group (EIG) owns a ridiculous number of web hosting companies (listed here). EIG does not have a great reputation for service quality and engages in aggressive upselling.8 While I expect most users will have a satisfactory experience with EIG-owned brands, I think most people can find better options at similar price points.
Well-known brands owned by EIG include Bluehost, HostGator, HostMonster, JustHost, FatCow, and iPage.9
Small or new web hosting companies
There are not a lot of barriers to entry in the web hosting business.10 It’s easy for individuals to start hosting businesses without necessarily having the commitment, skills, and/or desire to run a high-quality service for a long period of time.
Predicting future service quality is especially difficult with new web hosts. I feel good predicting that a company that’s had great uptime for the last decade will continue to have good uptime next month. On the other hand, if a new host has had great uptime over its six months of existence, I won’t feel as confident predicting its future uptime.
Warning: what follows is an in-the-weeds description of my methodology. Some readers may find it uninteresting or overly technical.
When evaluating web hosts, I consider performance data, prices, support quality, and a couple of other characteristics. I try to come up with a few different recommendations that are tailored to different situations a consumer might find himself or herself in. For example, I might recommend one company as a great bet for a consumer with a low-traffic website and a desire for solid performance on a limited budget.
I don’t arrive at my final recommendations in an algorithmic manner. For more on that, see my detailed blog post on issues with formulaic scoring systems.
Performance data collection
I have 12 test websites with popular and/or well-liked web hosts. The hosts, plans, and other details can be found in the table below.
|Host||Plan type||Plan name||Expected server location11||PHP version12||Initial purchase date|
|1&1 IONOS||Shared||Business||Wayne, PA||7.3||02/05/2019|
|A2||Shared||Lite||Ann Arbor, MI||7.3||02/04/2019|
|InMotion||Shared||Launch||Los Angeles, CA||7.3||02/05/2019|
|Namecheap||Shared||Stellar||Los Angeles, CA||7.3||02/05/2019|
I have usually selected the cheapest plan available from each host. If I was given an option during signup to select a specific data center location, I tried to choose the location closest to Chicago.
Test site configuration
On each test website, I’ve set up nearly-identical WordPress installations. Performance tests take place on the homepages of these websites which all contain a moderate amount of identical HTML and image content. The test websites run the default Twenty Nineteen theme on a homepage with roughly 1 megabyte of image content and a large amount of dummy text (equivalent to the length of roughly 20 typical paragraphs).14
I believe WordPress is a good application to use for testing purposes since (a) WordPress is a common content management system used by individuals and small businesses and (b) running WordPress involves a lot of processes that are typically used on non-WordPress websites.
The test websites serve text, HTML, images, and CSS while making use of PHP scripts and MySQL databases. I expect the performance of WordPress on a given host will generally be a decent proxy for the performance of other types of websites that involve similar processes (e.g., websites using the Joomla! or Drupal content management systems).
All of the test websites have .com domain names that I randomly generated and registered through Namecheap. At the moment, I’m keeping the domain names of the test websites private and using Namecheap’s WhoisGuard to prevent manipulation of the performance data.
When possible, I configured all test websites to use PHP 7.3. When hosts did not support PHP 7.3, I selected the most recent version available.
To create a level playing field for comparison, I have tried to disable server caching on all of the test sites. None of the test sites use content delivery networks.
I’m working to create a huge database of performance data assessing web hosts that cater to individuals and small businesses.
There are three different types of tests I regularly run:
- Uptime tests
- Page load speed tests
- LoadImpact tests
- Chicago, IL (the primary location)
- New York City, NY
- Los Angeles, CA
- Montreal, CA
- London, UK
- Hyderbad, IN
- Johannesburg, ZA
Page load speed tests
I also run page load speed tests using Site24x7. These tests assess the load time of web pages using a Firefox web browser.16 Load speed tests originate from the same eight locations that uptime tests originate from. A test originates from Chicago once every fifteen minutes and from other locations once every thirty minutes. The time it takes for different parts of the page load process (e.g., DNS time and download time) are recorded separately.
I use LoadImpact to assess how the test websites perform under heavy loads. I gradually simulate more and more virtual users to see how performance degrades as the load becomes more intense. I typically pause uptime checks and page load speed tests while LoadImpact tests are running.
Graphical results and commentary from my first round of LoadImpact tests can be found in this blog post. Please note that I’m still figuring out how I want to run these tests. I have limited confidence in some of my interpretations of the test results.
- I may be placed on an atypical server from any host (i.e., a server may have atypical hardware or an atypical shared environment relative to the typical server a host offers).
- Some servers will have an advantage because of their proximity to areas my speed tests originate from. In the future, I hope to aggregate speed data from multiple locations to reduce bias related to data center locations.
- Page load speeds are going to change when WordPress is updated or the configuration of my test sites changes. Comparing speed test results over long time periods may be misleading.
- The process may be biased towards hosts added at later dates (i.e., future hardware offerings may be better current hardware offerings—old accounts may remain on old server hardware).
- People often use caching or content delivery networks (CDNs). I don’t evaluate how those services affect the performance of websites on shared web hosts.
- In the future, I may not be able to maintain a level playing field across all hosts (e.g., I don’t expect to disable content delivery networks CDNs or caching on cloud services).
- Some hosts include SSL certs. Some SSLs will tend to have faster handshake times than others.
- Hosts could figure out which accounts belong to me and act to improve my test websites’ performance.17
- There is a lot of room for improvement in how the combined web hosting performance dashboard presents results from the underlying data.
Potential, future improvements
In the future, I may make adjustments or additions to my process for collecting performance data. Steps I’m considering include:
- Adding additional shared hosts.
- Including more of the larger web hosting market; evaluating managed WordPress hosts, cloud hosts, and/or low-cost VPNs.
- Adding multiple plans for hosts that claim to offer different performance levels on different plans.
- Creating a data tool that allows for custom queries.
- Creating test sites running an application other than WordPress.
- Creating test sites with caching and/or content delivery networks enabled.
- Testing how intense a server load I can generate before running up against TOS limitations.
Other factors considered
I pay close attention to how expensive web hosts are. I pay special attention to discrepancies between introductory prices and renewal prices. I see it as a positive thing when hosts allow users to sign up for short billing cycles (e.g., monthly) without paying substantial additional fees.
When thinking about prices, I also consider what hosts charge for add-ons that hosting customers may buy.
It’s becoming a norm for websites to use TSL/SSL certificates. I love to see hosts partnering with Let’s Encrypt to offer free, automated SSLs to their customers. Short of that, I like to see hosts offering low-price SSLs. When companies like GoDaddy push absurdly expensive SSLs on their customers, I get angry.18
It’s extremely common for hosts to try and upsell their customers on additional services. I’m ok with upselling in some scenarios, but a lot of the upselling in the hosting industry is distasteful. Some hosts are extremely aggressive about upselling, including pitches for multiple products in regular emails, in hosting control panels, and within sign-up forms. Some of the upselling is especially frustrating—many services are pitched in borderline dishonest ways (e.g., SiteLock).
It’s great to have easy access to knowledgeable support people. In my experience, support quality varies massively between hosts. At the moment, I think my evaluation of support quality is the weakest aspect of my evaluation process.
I draw on my own experiences with support agents, but that experience is limited and won’t necessarily be representative of the typical person’s experience.
At the moment, I have not thought of an easy or straightforward way to adequately assess support quality. Approaches I’ve seen other people use generally take the form of: “Ask live chat support agents dumb, easy questions and see how quickly they respond.” Tests like that are not useless, but I think what’s most important is how support agents handle real, hard problems. I’m not convinced that the speed at which live chat agents responds to basic questions is a good proxy for the overall quality of a web host’s support.
Since my own experience with support teams is limited, I draw on information about other people’s experience with web hosts’ support agents.
I draw on information about other people’s experiences with web hosts rather than relying exclusively on my data and my experiences. Unfortunately, it’s hard to find reliable sources for this information.
Even where legitimate reviews can be found, selection bias is likely to be an issue. For example, the typical user of GoDaddy’s services may be different from the typical user of a SiteGround’s services. Similarly, the typical individual who reviews a plan may be different from the typical user of a plan (i.e., I’m more likely to review a product when I love it or hate it than when I feel lukewarm about it).
The forum WebHostingTalk is probably my favorite resource for information on other users’ experiences (although I still think the experiences posted there need to be approached carefully and critically).
I like to see hosts offering feature-rich services. Most decent-sized, Linux hosts catering to individuals and small businesses offer the features typical users want to take advantage of (e.g., phpMyAdmin or automated script installers).
Since most features that are useful for individuals and small businesses (my target audience) are available across web hosts, I don’t display the sort of extensive feature lists and tables found on many other websites offering web host recommendations. However, I appreciate it when hosts offer atypical but useful features that cater to a broad audience (e.g., user-friendly staging tools for WordPress and Joomla! websites).
Ease of use
I like it when it’s easy to use a web host’s platform. I’m a big fan of the classic cPanel hosting control panel. cPanel is pretty easy to use, and since it’s common, the internet is full of tutorials explaining how to do things in cPanel. When hosts use other admin interfaces, I like to see that those interfaces are easy to understand and navigate.
I appreciate detailed, up-to-date knowledgebase articles. I also like to see signup processes that are simple and billing systems that are straightforward.
Organizational quality & competence
Data on performance and prices can tell me a lot about a host’s offerings and quality in the past. However, I can’t be sure that what I’ve seen in the past will necessarily continue on into the future. For this reason, I try to think about an organization’s general competence. Positive signals of the competence I look for include professional communication styles, transparency about business structures, and commitments to making the internet better or more secure.
- Prices and promotions change regularly, I won’t always be up-to-date.
- I don’t have a rigorous process for comparing support quality between web hosts.
- I only cover hosts catering to websites with fairly standard needs (e.g., I don’t evaluate Windows hosting platforms).
- I receive commissions from some hosts. I am neither impartial nor unbiased.
- Tech-savvy users may be better served by providers I don’t currently evaluate (e.g., DigitalOcean).
- As of 3/4/2019, Namecheap’s Stellar plan is available for $2.88 per month (archived source). For more price-related information, expand the “Plans and pricing” section below.
- Go to the web host performance dashboard and select “Namecheap dashboard” for more detailed information about the speed and reliability of Namecheap’s service.
A list of features offered on shared plans can be found on Namecheap’s shared hosting web page.
- Many popular hosts advertise amazing rates that only apply to the first billing cycle and require a multi-year commitment.
- As a general rule, the majority of websites are not particularly resource-intensive and have modest hosting needs. Namecheap’s service should be more than adequate for these websites.
- Go to the web host performance dashboard and select “SiteGround dashboard” to see more detailed information.
- For further details, see the section of my SiteGround review that covers support quality.
- On this archived page from 3/18/2019, you can see that (a) SiteGround is upfront about the fact that initial rates are only introductory, and (b) SiteGround clearly displays what prices customers will pay upon renewal.
- Users of the forum WebHostingTalk frequently have complaints about EIG’s services. I discuss upselling in The Web Hosting Confusopoly (details about my experience with EIG’s upselling can be found in one of the article’s footnotes).
- I have an extensive list of EIG brands on another web page.
- For example, many web hosts offer reseller packages that allow individuals to resell their services. As of 2/11/2019, HostGator has a reseller plan starting at just $19.95 for a month of service. An archived web page showing HostGator’s reseller offerings on 2/11/2019 can be viewed here.
- The recorded locations are based on what I came up with after looking up each server’s IP address then running the IP address through a location-finding service. It’s possible that these locations are incorrect or imprecise. Additionally, the locations may become inaccurate if a host migrates one of my accounts to a new data center.
- I aim to run the latest PHP version each host supports. If it looks like I’m not running the latest version available from a given host, please let me know.
- When ordering, I selected that I’d like the server to be at Hawk Host’s Dallas, TX location. My invoice from Hawk Host confirms that I made this selection. I don’t know why I ended up with a server in Amsterdam.
- While the test sites have a fair amount of text, the text contributes little to the overall page size in terms of bytes.
- Although these tests are set to run once per minue, they may in practice run closer to once every two minutes.
- “Render your website using a real browser (Firefox) in real-time.”
From Web Page Speed (Browser) | Site24x7 (archived copy). Accessed 1/15/2019.
- At this time, I expect that this kind of manipulation is extremely unlikely. If my web site becomes more popular, the risk of manipulation may increase.
- The cheapest SSL I could find on GoDaddy’s website on 2/11/2019 was $74.99 per year ($67.49 for the first year). An archived copy of GoDaddy’s web page with SSL offerings on 2/11/2019 can be viewed here.