November 29, 2012

Why Your Slow Server Is Killing Your Site and How To Fix It

 

GeekRev.com is not, by any means, one of those blessed sites that get millions of hits a month. But, our traffic levels are pretty good, and very “qualified.”

In marketing, “qualified” sources refers to those sources that specifically come to your site on purpose. There are a lot of ways traffic becomes qualified. For instance, someone can specifically come to your site from a business card, word of mouth, or something similar, because they know for sure that your site has the information they need. Qualified traffic can also come from very specific search-engine queries. Most people come to this site on purpose.

But a few months ago, I noticed our traffic was quickly declining. Something was wrong and I was not sure what it was.

Houston, we have a problem.

A couple of things came to mind. First, I read a lot of blog posts saying that Google’s search parameters had changed again, and that many sites could be in danger. But after comparing GeekRev.com against the list of newly forbidden authoring “techniques” I realized we violated none.

A second possible culprit was an issue with the server. I checked our logs for errors, and nothing really seemed outside of the norm. I was very confused. After a couple weeks of watching key metrics getting worse (bounce rate, pages per visit, time on site), I chocked it up to Google search idiosyncrasies, and decided to wait out the storm hoping Google’s next batch of changes would get things normalized.

Things got worse.

For instance, my bounce rate, previously an impressive 4%-5% (yes that’s right) was creeping upwards, hitting 10%. While most sites are healthy at 30%-50%, GeekRev’s historically high rate of qualified visitors gave us an uncharacteristically low bounce-rate. So, this jump to 10%, while still better than most sites, is obviously alarming.

I was getting frustrated trying to figure out why everything was getting so bad. Finally, it got the best of me and I just decided to turn off the Google Analytics widget in my site’s dashboard and ignore it all, hoping things will naturally get better. But that did not happen. Things got worse.

Over time, even without looking at Google Analytics, I saw the traffic to my site dropping, to about 60% below normal levels. I also noticed something else – my server seemed to get slower and slower. But the difference from day to day seemed statistically insignificant. I thought maybe I was letting my imagination get the best of me.

I started following the clues.

I was cleaning out some image files when I came across a screenshot I had taken of my site’s load speed from Pingdom. It showed a load speed of 3.85 seconds. Pretty good. Out of curiosity, I tested the site again. 15.21 seconds. What??? That can’t be right. Test after test confirmed load times in the neighborhood of 15 seconds. Could I have found the problem?

I remember reading somewhere that a typical desktop web user gives up if a site takes longer than seven seconds to load, and the average mobile user gives up after 10 seconds. I think the extended patience of a mobile user is because mobile speeds are typically slower than one expects on a desktop, so mobile users are accustomed to waiting for sites to load.

If that’s right, I was losing at least half of every mobile and desktop user who came to GeekRev.com. Ouch.

Curiosity set in and I checked out the Google Analytics page. My bounce-rate was now 80%. Remember, it used to be 5% at the most. All the other important metrics were similarly terrible – except unique visitors (the number of individual people attempting to visit your site). So, just as many people were coming to the site, but they were *not* staying, they were *not* clicking more pages, and they were *not* staying on pages very long.

I developed a hypothesis.

While changes at Google may still have had an impact, since I had not seen much real growth in the unique visitors statistic (stayed relatively level), I reasoned that most people where actually clicking to visit GeekRev (highly qualified as usual). But, when the page was taking 15 seconds or more to load, they just gave up and left. That gave me a “visit” hit that turned into a “bounce” because they just left. 80% of the people coming to my site were leaving before seeing a single bit of content. My stomach began to ache.

Could it be that the server speed was having that great of an impact on my site’s metrics?

I decided to act.

Armed with this hypothesis, I decided to move the whole site to a hosting provider I had used in the past, who I knew had a great speed rating. Of course, it was going to cost more to host GeekRev now, but I figured a few months “testing” should let me know if the cost would be worth the benefit.

I’ll make a new post later on that details exactly what I did to make the move. But, for now, I’ll just say that I copied all the content to the new server, but left things intact on the old server (just in case I wanted to move back). Then, I changed the Domain Name Server (DNS) settings to direct people to the new server.

I could not believe the difference.

Right off the bat I could tell the new server was much faster. Pingdom says the site is loading between 2.5 and 3.9 seconds. WOW! Very fast! But did it fix the problems? Well, let’s look at some charts.

This image is the bounce-rate graph from Google Analytics. The star indicates the last day the old server was live. The following day, while the DNS was still propagating (which means some people were still seeing the old site), the bounce rate dropped to 40%. The first full day on the server, the bounce-rate dropped to 0%! This past Saturday (the day highlighted) you can see it only got to 1.37%! That is freaking amazing!

What about the other metrics?

This image is the pages per visit graph from Google Analytics. Again, the star indicates the last day the old server was live. This metric indicates how many pages, on average, individual users visited while at my site. It has steadily, and rapidly increased since the new server went live, and it is set to double in a couple of days!

This image is the page views graph from Google Analytics. Again, the star indicates the last day the old server was live. This metric indicates how many pages are being viewed per day. Again, there is a lot of growth! Those extra pages per visit are translating into three times the number of pages being viewed as before the move. That means more content is being seen, more ads are being served, and my site looks better and better to potential partners.

Bottom-line: a faster server matters.

These charts are showing dramatic results from the server move. I am sure some statistician could dampen my results in terms of how much improvement the average site could expect from ditching their slow server. But for me, for GeekRev.com, it is plainly obvious that the slow server was a hidden villain that was murdering my site’s stats and destroying my visitors’ user-experience.

I do want to say there is plenty you can do to try to speed-up your site without making such a drastic move. Optimizing images, compressing files, and caching are a few such techniques. Try those first. Then, if you still find yourself with a slow server, there is really nothing you can do about it but to move to a faster server.

Have you seen similar results? Let me know in the comments!

CC Image • JohnSeb on Flickr.com

Originally Published: November 29, 2012
Category: Technology
Sign Up Today. Membership is free

Public Stream