Seven Deadly Web Analytics Sins    1    2    3    4    5    6    7

By John Marshall

'People commonly use statistics like a drunk uses a lamp post; For support rather than illumination' -Mark Twain

Mark Twain probably didn't have web analytics in mind when he made the above statement—but he might as well have. Far too often, the purpose of web analytics is to produce a graph that goes up and to the right. But, good, meaningful web analytics is a deductive process; not just an effort to produce a chart.

So how do you keep from sinning during your web analysis? By investing the time in truly exploring the data, gaining insight into its cause and effect and understanding the underlying data points while simultaneously filtering out the noise.

Seven Deadly Web Analytics Sins is a series of articles; each article unveils another web analytics sin. We start with: "The sin of simple visitor counts."

Deadly Web Analytics Sin: Simple Visitor Counts
All visitors are NOT created equal. Many entities arriving at your web site may appear to be human visitors, but in reality are anything but. Unfortunately this inhuman traffic is counted along side your human visitors—so you can see why 'number of visitors' data can be faulty. A few of the factors that lead to non-human visitors include:

Bad Bots
Bots are robots that visit your site to extract data from its pages. 'Good' bots include search engine crawlers like Googlebot and Yahoo! Slurp. You definitely want good bots to visit your site so your pages will show up in search engine results.

Good bots follow protocol, identifying themselves to web analytics programs so that they aren't mistaken for human activity. But bad bots don't play by the rules. Out to harvest data like email addresses they don't announce themselves, preferring the cloak of anonymity that masquerading as a human visitor gives them. The bad bots know if they ID themselves for what they really are, webmasters would simply block the bots and reduce the shady activities. So they simulate human behavior and end up skewing your stats.

Fraudulent Clicks
A very real, rapidly growing problem, click fraud occurs when someone (or some computer script) knowingly clicks on your PPC ad with no interest in your products or services. Whether they're trying to fraudulently make money from the traffic being sent to your site, or are maliciously attacking your ad budget, one thing is clear--when click fraud occurs, your wallet deflates while your stats inflate.

Dead or Alive
Uptime monitoring services are 'good' traffic in that they provide a useful service. By probing your site as if they're human visitors, they alert you to problems human visitors might be experiencing. But useful they may be, they're never going to buy your widgets. They will, however, skew your data.

The Harsh Reality
Any of the above examples can cause a spike in a simple visitor count or an upward trend. Getting past this bogus information and to the actual human behavior requires a willingness to examine the data and not just take comfort in the stats, however uncomfortable that may be. Looking at a variety of stats in conjunction with visitor counts helps you put this fallible stat into perspective.

The Path to Righteousness
A metric must be defined to bring into focus only 'qualified' visitors. Different sites have different methods of defining this, but a fairly universal method is to include only those visitors that spend more than 10 seconds on the site.

This will exclude many bots and instances of click fraud because those entities are likely to visit only a single page. In web analytics we can only measure time on page/site at the second click, since it's the second click that actually defines the boundaries of the first page time period. (For example, the program doesn't know when someone closes the browser window.) It's worth noting, however, that the 10-second-or-more label may inadvertently exclude some small number of real human clicks, like those who click through from an ad, don't like what they see on the single landing page and immediately exit. That's just the way it works.

Given that the 'number of visitors' is a fairly vague metric for defining business success, it doesn't matter too much that we exclude some real human traffic. Especially since we're interested in only the 'qualified' visitors—and a visitor who clicks in and immediately out of your site wouldn't even be considered a lukewarm lead by most!

The practice of exclusion will obviously reduce your metrics, perhaps causing to you tremble at the thought of going to the boss with graphs that are lower than last month. Oh no! What shall we do? Why not e-mail a copy of this article to your boss before showing the new (and improved) data!

Note: ClickTracks' software contains algorithms to help distinguish bot activity from human activity, even when the bot does not politely announce itself. This helps reduce the impact of bots on a simple visitor count, but click fraud and other non-human activity remains a significant volume of traffic for many sites.

Seven Deadly Web Analytics Sins    1    2    3    4    5    6    7