Buy Now

Hypothesis Test Backstory

by Philip Mayfield

In February of 2011 I received an email from an irritated customer which indicated that my website ( was very slow. In limited cases I had previously observed the same, but assumed it had something to do with my computer. The most obvious solution to a slow site would be to notify my web hosting company, Alentus. I did some preliminary testing and it did appear that my site was not only slower than big sites such as Google and Corel, but was actually much slower than Alentus' home page.  I was more than a little irked that the company I was paying for hosting, Alentus, had a much faster load time than my site.


My initial data set is below. I should point out this is not the data used in the following article. This sample was much too small and was augmented later. The key number in the table below is "Average Speed per KB" and the lower this number the better. By using the average speed per KB, we provide a statistic that adjusts for the differences in web page size.


Original Hosting Times


I sent this data to my hosting company asking if there were a way to increase the speed of my site without changing hosts. The short version of their reply was that I either needed to change to dedicated hosting, which is much more expensive, or change servers. At this point, I had two options…

Option A: Do Nothing

Option B: Do Something (change hosts, change servers, use dedicated hosting, ….)

If I chose to do nothing and my site was truly slow, then my customers would suffer. If I chose to do something, but in reality my site was not slow, then I would be expending a lot of effort which wasn’t needed. At this point you might think that I had sufficient data to act. To act on a single data point would be folly. The differences in speed could have been due to my local internet connection, the day or time of day I ran the test (maybe Alentus was doing a server update), or perhaps Alentus was under a Denial of Service attack. I needed more data.


My company happens to have another website with HostMonster which we use for FTP access. So, I loaded three pages that were hosted on Alentus to HostMonster. The three pages were... (home page)


I had three different employees collect page load times over a period of 2 weeks using a site called WebWait. You can download the website hosting comparison data set here.

Click here to go back to the article.

Click here to learn more about Quantum XL



dividerleave us a comment