Content Under Consternation

Right off the bat, I'll admit I'm a bit irritated that one of the most popular and innocent pages on this web site, Natural Angels (a simple, naive poem), happens to share a common title with less innocent and more-spine curving matches on Google. As much as I like to yammer about the marketability, layout, and user behavior with Web-based content, I haven't done a great job of promoting my own content. Sometimes I have an almost sick fascination with not promoting any particular content and watching what search terms people use to locate certain pages.

Over the summer, I added a basic five-star rating system to my content pages. You click a star to denote your level of satisfaction. For the most part, people ignore it. Maybe it's not clear that anyone can contribute a rating. So people don't have to fuss around with registration and I don't have to worry about long term session ids, the ratings are stored by IP address.

Recently, I've been growing more interested in looking at the behavioral trends of how groups of people, or everyone, interact with content over time. There are some obvious take-aways, such as people gravitating to the scrollbar on the brochure page in part because of the blog entries. Other pages show unexpected patterns, such as motion concentrations and movement originations from the upper left hand corner.

Also, I've been making small changes to the RDF descriptors in my Creative Commons licensed content to try and increase search engine exposure. The other day, I reviewed the query referals for several URLs linked by various search engines for the first few days in December. One of the queries was for a geometric value, and Google directed this particular person to the science fiction story, The Alchemy of The Aurora Chateau Deo Belle Etoile. I remembered the query entry because it stood out as an odd match for the content.

Today, I was uploading some new indices and new content, and I checked the modified dates on the comment files, which contain the ratings for each page. I noticed the comment file for Alchemy had been updated, and I quickly checked. And I sat back, staring at what amounted to be a slap-in-the-face. Two stars. I always thought that Alchemy was a fun, tongue-in-cheek story. Not a five-star story, but at least a three-star story. Sure, there is a lot of exposition and it concludes with pretentious drivel, but it is a fun story. I never thought of it as a two-star story, and I know I have more than one that qualify as, at most, two-star stories.

One of my biggest gripes about online rating systems is that they are too anonymous and too blind. Someone votes you or your work up or down, left or right, and you have no insight into what was really going on. Worse, there is always someone or something of lesser or inferior quality that garners more and higher ratings. I don't think of it as whether only a subset of people should cast a vote, but I do think that if people are voting in an unexpected pattern or voting certain content or items higher than others, it requires further investigation. Not that I necessarily want this particular company but I think Christine Gregoire, John Kerry, and Al Gore would agree with that sentiment. What I need in my case is an exit poll for content. Oh, wait, I already have one.

One person, the only person to vote, gave Alchemy a two-star rating. Well, maybe they didn't like the story. I have to concede that. Although, they were searching for terms - something about geometry and pyramids - that indicate a humorous science fiction story was probably not the desired destination. But, maybe they found the story, took the time to read it, decided they didn't like it, and gave it a two-star vote.

Let's check. The general motion indicates this particular user didn't travel very far on the page. And, believe it or not, you generate a lot of motion activity even if you don't use your pointing device. I designed the behavior monitor to create junctures between motion and interaction, so even if you make no motion at all, any interaction automatically records an entry for motion. Why? Precisely because the interaction is motion, just not with the pointing device. Typically, such motion is concentrated in one area. Ok, so this particular entry didn't generate much motion. Let's check the interaction (height was clipped). Yes, there they are, voting for two stars. A bit hard to read from this view. The interaction points are: clicked star #5 at 15.0 seconds, clicked star #5 again at 16.0 seconds, and then clicked star #2 at 26.0 seconds. Eleven seconds to decide between five stars and two stars. Finally, let's check the track (height was clipped). You'll notice that the first twenty-six seconds of the total thirty-one second visit was spent fussing with the rating bar, with the eleven-second delay between the five-star and two-star rating. An eleven-second decision. Given the search criteria, the motion concentration, the time spent interacting with the rating bar and the little time interacting with the content, I think this rating probably should not have been given, or at least shouldn't have the same weight as someone who read the story and then decided it was crap. I'll leave the two-star rating in place for now, because I don't want to be accused of tampering with my electronic voting system.

For comparison, I have a number of examples with limited track movement and motion concentration. In this case, the motion pattern and track timing indicates this person spent more time with the content. Also, the interaction report confirms the interaction, as opposed to the viewer simply opening up the page, hanging around for a while, and then leaving without doing anything. If this particular user gave the content a low rating, I could argue more gray and straw-grasping points such as the user didn't understand my complex humor. At least I would know the reader spent time with the content; four minutes, nineteen seconds to be exact. Anyway, I have found many interesting motion and track patterns that provide more meaningful examples of how people interact with Web pages.

It's very important to understand how people interact with content. Recently, I was on the phone with Gunter Sushi, trying to explain how a particular group of users were having a terrible experience with a certain Web site. Gunter was of the opinion that the data was wrong, that no one was having such a ridiculously poor experience. I calmly explained how the data was sampled, offered to show a simple repro, and pointed out where the specific type of user was having the problem and why it was so bad. I thought it was a very clear presentation of undeniable facts. Yet, Gunter denied each and every one of them. At first the facts were argued, and when those facts were proven, the clarity of presentation was argued. When the presentation was reduced to the most simple and pertinent points, the marketability of those points was argued. Even if those factoids were true, how could those factoids be presented in such a way as to make the necessary corrections? At that point, I had to tell Gunter that neither my software nor I was there to manage his employees. I'm using this as an isolated instance, but I have seen this situation repeated many times, unfortunately in direct proportion to the amount of revenue generated from the Web site. The more money it makes, the less people understand the innerworkings of their own site. While Gunter Sushi works up the courage to tell his employees that they have a bug in their poorly written code, a significant number of people are unable to make purchases. This wasn't brain surgery. We're talking about some pretty cut-and-dry Web design using arcane API's.

For some reason, discussing users interacting with content perks up people's ears, but discussing users interacting with their content is taboo. What is ridiculous about this is that some of these large sites are loaded to the hilt with monitors. They've got back-end monitors and front-end monitors, active and passive monitors, hit counters and marketing trackers, pathing and trend analysis tools, and log shredders up the hoo-ha. And for all of the available monitoring solutions, many of which get piled on top of each other because internal business and development groups can't agree on what they need, most customers still have absolutely no idea what takes place between a page loading and a page unloading. While path analysis tools (aka: click stream) can determine where a visitor left the site, most products can't tell why, and almost no product will tell you what happened. Which is what makes conversations with Gunter Sushi and his likes so invigorating. On the one hand, they have designers telling them what looks good, and they have developers telling them what works well, and they have marketing analysists telling them their advertising campaigns are successful based on the rate of return. On the other hand, they have someone like me who comes in and says, it may look nice, it may run well, and all of your monitoring solutions may sing your network praises, but this and that type and percentage of users are having a terrible experience. How do I know? Well, the results clearly show a particular body and class of visitors are having trouble. And before they interject, I have to add quickly, I realize you are not receiving any error reports on your server. This problem is not on the server, it is a problem with the content that affects a small percentage of users.

You probaly already know the typical response: That's not our problem. We only need to know if the network is slow or inaccessible. I understand that certain roles and responsibilities belong with certain groups. But user interaction and user behavior with content seems to be often times entirely overlooked. So, who owns this problem? For example, one way to look at it is one-half percent of all traffic is simply part of the overal non-converted customer traffic, or one-half percent of all traffic is prevented from being converted into a customer. Shouldn't that be a concern? For many enterprise web sites, it's not even on the radar. I don't understand this. If you make one million euros (that's my localization support for the day) a month, that's five thousand euros. I'm not suggesting a direct conversion, but still that is potential revenue being tossed aside.

For some reason people seem afraid to consider how users experience their Web sites. Not too disimilar from the two-star rating of my Alchemy story. I want to see what people think about my stories, while I am also uncertain about whether I really want the criticism. Isn't it easier to brush that two-star rating under the carpet because, as previously mentioned, the user probably wasn't interested? Or, should I take that two-star rating and make some extrapolation. Say, for instance, nothing on the page caught the viewer's attention. The reports prove as much; the viewer was more interested in the rating bar than the rest of the page. If I had a more compelling element on my Web site than the current contents, and millions of visitors flocked to see it, would that change whether or not the layout is good or bad, the color scheme is pleasing or nauseating, script errors are being thrown left and right, or all I see is a big plugin icon because the whole site was created in flash?

I'm not selling anything here. But, if I were, I know that, for the most part, the site shows up okay in most browsers. I really don't care how it looks on a cell phone or Lynx. But, I was pretty happy when I checked out this site with Safara for the first time since adding the calendar, and seeing that work on a Mac brought a little smile to my face. I knew it should work since I tested it with Konqueror; just certain XML parts for engine don't work. And opera mostly works. Why am I happy about that? Because I didn't design it around particular browser APIs, and I didn't really have to change anything. I only made sure it worked in Mozilla/Firefox and Internet Explorer, and everything else fell into place. And, unless I'm missing something obvious (besides my routinely non-validating XHTML Strict declaration), there aren't any errors that impede using this site. And, depending on the page, there is quite a bit of background scripting. So, it's not hard to have a web site that, most of the time, works without errors on most browsers.

Why is it so hard to recognize that a part of Web site experience is just that: understanding what the user is experiencing on the site, and not what it looks like from the server point of view? Counting log hits and ratings won't tell the full, unvarnished reality about how a user experiences your web site.