I'm a web analytics junkie. I admit it. I use no less than 4 analytics programs. One is a log analysis program provided by my hosting provider, one is a log analysis program I run locally, and two are tag-based, hosted analytics programs. I just got my invite for the free Google Analytics program (formerly Urchin) which is partly what inspired this article.

If you are interested in web customer experience, you should become a web analytics junkie too. The data you can access are a primary, continuous source of feedback that you can use to improve your performance. The key is to understand what you need to pay attention to and what you can afford to ignore. You need to figure this out for yourself primarily because every situation is unique. Focusing on these broad issues is beyond the scope of this article, but there are a number of books, websites and forums that can help you. My intention is simply to stress the importance of obtaining, analyzing and acting on this information.

The presentation of data is not a new field. However, I think it is useful to occasionally re-visit well-worn topics for the benefit of those who may not have been around when we originally beat these horses to death. I don't want to go into the entire, broad topic of data presentation. Volumes have been written on the subject. A good place to start would be Edward Tufte's Visual Display of Quantitative Information. It may sound boring, but it really isn't. Just about anything written by Tufte is worth a look in my experience.

I want to talk about pie charts primarily. If you are using pie charts on your web site, in your PowerPoint presentations, in your corporate documents, etc....STOP. Just stop.

There are really two pieces of information you are typically trying to convey by graphically representing data. Either you want to communicate the differences among values (or proportions or whatever) or you want to communicate the absolute values of each alternative. The alternatives presentation styles available to you will almost always handle these situations better than pie charts. Let's do a simple experiment.

First, I'm going to show you a pie chart representing 3 values. Time yourself and record how much time it takes for you to rank order the choices from highest to lowest value. Ready? Go!

OK. Now we are going to do the same thing with the exact same data, but with a different type of chart. Time yourself again in ranking the values from highest to lowest. Ready? Go!

If your brain works like 99.999% of the rest of us, you were able to rank the bar graph much more quickly than the pie chart. I used an extreme example to make a point. But your mind had no problem ranking the second chart. Some of you may have thought that the pie chart all had the same values, but I assure you that the data is exactly the same.

You could do the same sort of test with the actual values. Even if you labeled the pie chart, your comprehension will be quicker with another type of display of the data most likely.

With this in mind, I'm still surprised to see so many analytics companies relying so heavily on pie charts. To my mind, the best use of a pie chart is when you have one value that is overwhelmingly larger than the rest and you don't want the audience to focus on the actual values, but just bamboozle them with the overwhelming size of the leading segment. Of course, this seems to come close to embracing the old adage, "There are lies, damn lies and statistics."

Of course, if you need a statistic quickly, you could just make it up. According to what I heard, over 79% of statistics are made up on the spot. But that isn't very compelling, is it? Now you can add a little credibility to your made up statistics using Philipp Lenssen's excellent eDataSolutions (http://www.esolutionsdata.com/). Just type in your stats, create your conclusion and click. You get a publicly accessible web page to support your bogus numbers. Pure genius. A complete sham, but genius all the same.

I want to talk about one other issue in data visualization. I came across a web analytics package called VisitorVille (http://www.visitorville.com/). I'm sure that every reviewer who comments on this application calls it a combination of web analytics and The Sims video game. Basically, it turns your web user data into this virtual environment where users are represented as different types of vehicles or little animated people and your web pages are represented as locations around a town. While this has a serious "Gee Whiz!" factor, I can't really see this type of information visualization breaking through to the mainstream. The meaning of the information being represented just isn't intuitively obvious enough. You have to insert this simulated environment metaphor between the data and your understanding (and any actionable information). Here's a screenshot:

I should probably make it abundantly clear that I have not used this product at all. I'm simply raising a general issue based on the information available on their web site. If it were free, I'd certainly try it out just to see what it is all about. But the new way of representing data that this product offers is not compelling enough to make me commit to this product by itself. In other words, the meaning of the data is not intuitively obvious when presented in this manner. I'll take a table of numbers or a plain old bar chart every time.

Technorati Tags: Information Visualization, Pie Charts, Web Analytics, Edward Tufte, Google Analytics, VisitorVille