Fig.1: larger, Klong
Attention is a valuable resource and we better decide carefully what is worth spending it on and what is not. There are thousands of distractions of varying quality, both regarding the content and the presentation. If they are commercial, though, they all have one thing in common:
They try to bind your attention as long as possible.
Attention has become a currency, the consumer has become a product. Binding your attention is a business model. The "content" industry tries all kinds of tricks to keep your attention focused on their services. How does this fight for our attention form our habits? Let's have a look.
When people visit your home page (or whatever you have in its place today), you would expect the time that the visitors stay on your home page to follow a normal distribution. The corresponding scatter plot of visit lengths (fig.1) shows a dense cluster in the middle (around the mean length) and rapidly thins out towards the borders. The histogram (fig.2) of lengths would then exhibit a rough approximation of the expected bell shape [in this case X~N(147;492)]
This picture would probably reflect reality pretty well, if people could allot their attention freely, i.e. if attention was an unlimited resource. It actually is an unlimited resource when you are bored, so above distribution was what you probably could observe in the beginnings of the world wide web, simply because each discovery was cherished and every web site explored more or less thoroughly.
Here is an actual data set that I have extracted from the log file of my web server. The length of a visit is estimated by counting the clicks of a visitor and approximating the time between clicks (especially for zero-click visitors).
The distribution obviously does not look like the expected one, so what is happening here? Take body weight for an example. If there is plenty of food, food intake is mostly a matter of preference and other factors unrelated to the availability of food, so the distribution of body weights will be normal.
If, however, food was a limited resource, we would expect people to compete for food, so there would be a few winners who would be rather fat and then a huge mass (sic!) of rather thin people. This is because each time a winner takes a big chunk, the probability for another winner sinks significantly, because a limited resource becomes even more limited.
In the lognormal distribution (fig.4), the mass of small fragments (thin people, people with little attention to spend) is in the head of the distribution (to the left) and the large fragments are distributed accross the long thin tail of the distribution.
This explains neatly what happens here. Each time you get trapped in an attention sink, a large fragment of time will be lost, so the probability for studying something else equally thoroughly sinks, simply because there is not enough time any longer. When plotting the histogram of lengths, this time the result is a lognormal curve [fig.4; X~Lognorm(3.9;4.272)].
The distribution of visit lengths predicts that 10% of the visitors will leave the home page after a little more than a second, 25% after less than 7 seconds, and the average visitor will spend about 50 seconds. Here is a comparison of expected versus observed visit lengths:
| Percent of Visitors |
Leave within (s) |
Expected (s) |
|---|---|---|
| 10 | 1.2 | 84.2 |
| 25 | 6.9 | 113.9 |
| 50 | 49.4 | 147.0 |
| 75 | 354.6 | 180.1 |
(The last row of the table is where the model breaks down. Because the lognormal distribution has a very long, very thin tail, it does not reflect the bounded nature of a visit well.)
The people leaving within the first few seconds are probably those who
Of those who stay, the mean length of the visit is less than a minute, hardly enough time to study anything of even moderate complexity. (And then the mean is probably even overestimated, because there is an upper bound to visit length, which is not inherent in the distribution.)
Another effect of the lognormal distribution and the associated fragmentation of attention should be a quick turnaround on social web sites, because many people will scan the sites repeatedly for new distractions. For example, when submitting content to a news aggregator, like Reddit or Hacker News, most of the visitors should appear soon after posting the content. Let's have a look at some more data from the web log. (fig.5)
Indeed, after posting a link, 22.5% of the visitors request the submitted page within one hour, 48.5% within four hours, 74% within 10 hours, and then visitor frequency slowly returns to pre-posting levels.
Yet another interesting data set is the number of follow-up clicks per visitor. In a multi-page document, how many visitors read the second page, how many the third page, etc? The article examined in fig.6 consists of seven pages, allowing six follow-up clicks at most.
84.5% of the visitors do not even make it past the summary, which probably indicates that they were not really interested in the subject to begin with. 92.3% do not get past page 2, and then things start to level out. Almost everybody who read page 5 (4 clicks) also read the rest of the article.
It is not coincidence that we "pay" attention. Your attention is a limited resource and its is a good idea to spend it wisely. Our current culture fosters habits of fragmented attention and permanent distraction. Here is some food for thought:
Being the master of your own attention is more and more becoming a luxury, but it does not have to be that way. Choose the focus of your attention carefully! Really paying attention to something or someone can be quite a rewarding experience. Do not let it become a lost art!
