Analyzing Traffic Stats
While I was venting about short-term strangeness in my stats, it got me to thinking about how to read your stats and make sense of what is going on with your website. It is amazing the amount of data that is contained within your server logs, and with a good program (I swear by WebTrends) you can find out almost anything about your site traffic and your visitors.
One of the big things I check is the country information. You can learn so much from this set of stats. I remember once in the mainstream world, the site I worked for had a massive spike one day and we could not for the life of us figure out why. I mean this spike was huge, we went from a daily average of about 200,000 uniques to almost 520,000 from one day to the next. This was one of my first lessons in reading the stats numbers. Going through them, there seemed to be 2 major increases country-wise, for the US and for Hong Kong.
After a week of trying to figure it out, it turned out that there were two simple reasons. One was a press release issued by one of our partners in Hong Kong announcing the agreement and the second was another press release that we had issued a while before being printed in the US media. Both just happened on the same day. Since we had not seen either of these and had no advance notice that either would come out that day, it took us by surprise. But, there was a core reason and I found it in the stats.
There are other easy to find reasons for drops in traffic as well. Some of them are well-known and industry-wide, such as the traditional summer slowdown when people start heading outside to enjoy the nice weather. Kids are out of school and go online for games and music rather than homework, so the traffic differentiation between genres of sites changes. Overall volume may stay steady, but patterns change. Though, there is almost always a slowdown in summer.
But, when winter approaches and Christmas gets near, there is usually an increase, with more people going online to look for gifts for friends and family. This is another example of demographics shifting, as the shopping sites tend to get the majority of this holiday traffic. And with many people off for the holidays, they spend time online since they have spare time to play around.
Then there are the decreases which are actually good. How is this, you may ask. Well, at a few sites I have worked with, I work and work to build and build the traffic, getting it higher and higher and then suddenly it drops. Not a lot, mind you, but if I had seen growth at 10% per month, suddenly it drops a month or two's growth almost overnight. Why? AOL.
Once you get to a certain level of usage among AOL users, the company caches your site to speed up their service. So, many people start to get the cached version instead of the live version from the server, thus distorting the stats. There is an easy way to see this, just look for a large increase in traffic from the state of Virginia, where AOL has their main internet hub. This number will grow and grow and then suddenly drop, indicating that AOL has cached your site. I always looked at that as having "made it" knowing that AOL considers your site to be important and busy enough to cache.
You can get around this effect somewhat (as many sites are wont to do, since they have advertisers to please and ad impressions to sell) by using some simple code in your pages to ask servers not to cache your site - which can usually help alleviate this problem a little bit. These trends will allow for business planning and capitalization on Internet usage...
Search engine spiders can distort stats as well, either by simply adding traffic that is not real, or by really buggering up the numbers. I once worked at a company where we had some really weird days for stats. It seems that there was some sort of strange loop buried in the code of one of their sites that would literally trap spiders on the site. This particular site would go from 10,000 page views one day to over 800,000 the next - all because the spider got trapped and bounced between a couple of pages for the entire day. That little snag took the programmers almost 6 months to find, but it was easy to correct the problem once they found it.
And then there are the intended results of the spidering, where your site gets re-indexed and goes up or down in the search result rankings. Thus, you could see either a drop or increase in your stats for what seemed to be no apparent reason. If it works out that a few of the major engines re-index within a few days of each other, you could see serious changes in traffic flow. But, a good look at your logs will show the increase or decrease in clicks from certain search engines, making it fairly easy to figure out.
It is comforting to know that the longer-term traffic behaviors are usually not that hard to figure out. Especially as you extrapolate out into longer and longer time frames. These trends will allow for business planning and capitalization on Internet usage, which makes for more success. Mind you, the day-to-day ones still wig me out a bit, though. But, there is nothing that can be done about them, so I am back to my far-sighted outlook.
So, work with the stats you can, take the long term view and use the available information to capitalize on any trends that you see. That is the path to success.