When Do People Read Blogs ?
    By Tommaso Dorigo | April 15th 2010 01:42 AM | 13 comments | Print | E-mail | Track Comments
    About Tommaso

    I am an experimental particle physicist working with the CMS experiment at CERN. In my spare time I play chess, abuse the piano, and aim my dobson...

    View Tommaso's Profile
    Have a look at the figure on the left. It shows the number of visits to this site broken down in hours of the day -the time of the server used by the visitor. The statistics of each bar is sufficient that the uncertainty on their height is of the order of 2%, so almost indistinguishable by eye. What you can see, therefore, are real variations with time of the traffic to this site, and not random fluctuations up and down.

    You immediately note the night dip: people stop reading this blog after midnight, and resume after 8AM. Also note its asymmetric nature: the decrease is exponential, with a "hardcore" bottom of sleepless visitors; while the increase after 6AM shows a typical turn-on shape, the one you get if you integrate a Gaussian shape. It reflects the higher and higher active users as the morning progresses, when the time of start of their operations has a Gaussian shape.

    But there is much more to note. For instance, the dip at 2PM: it is significant, and evidences the lunch break that some take at around that time. Even more significant is the fall after 5PM, which is due to people leaving office and eventually having dinner. Activities increase again after 9PM, to then drop as night falls.

    I think this graph -or similar ones one might draw with large-traffic sites not connected to business- carries some sociological implications. It clearly demonstrates that people spend their working hours on the internet, more than their ours off work. It also gives an idea of the fraction of night browsers, which is larger than I thought. Finally, but this is a less-well founded hypothesis, the decrease of visits from 10AM to 5PM evidences our diminishing attention span as the day progresses.

    Do you have similar plots to compare ? I bet there are internet sites which devote a lot of attention to these issues. If you have a good link to share, let me know.


    A friend of mine studied this stuff for his PhD thesis in computer science. He studied the hourly breakdown for each time zone. Unsurprisingly, the peak is around 9am for each time zone. If you have a finer binning, you should see a peak in the first 10 minutes of every hour (bed ->go to work -> check emails -> check blogs)

    Why the first 10' of each hour Riccardo ? Maybe automated search spiders and such ?

    It is probably the time one spends checking emails and blogs before start working.

    I was not clear Riccardo, I questioned that such 10' peaks at the start of each hour cannot be explained by start of work. Only the ones at 8 and 9am, and perhaps 3pm. Anyway I have no finer binninv to check... Cheers T.
    Well observed, Tommaso.

    Before the age of home computers there were graphs like this from utilities companies showing domestic trends.  The Wembley cup final and the Aintree Grand National for example, would show huge spikes as people switched on their televisions.  Other spikes would show kettles being switched on during breaks.

    I had an idea for dealing with adverts that addressed adults like children: you switch off, not the TV, but everything.  After 1 minute, you switch on again.  Unfortunately, the reality is that a few million people doing this would damage equipment at the power station.  Such a pity.  Britain could have had the world's first electrical elections. :-)
    I enjoyed this post (this slight deviation from exponential fall during the night has to be me); I think Riccardo's explanation applies also to "the decrease of visits from 10AM to 5PM", which in my opinion doesn't "evidence our diminishing attention span as the day progresses" but quite the opposite. It's the slight increase after 3PM actually that evidences the diminishing attention span. (Unless someone's profession involves reading blogs; if yes, where do we subscribe?)

    What time zone are these in? Or are they adjusted to the time zone of the visitor?

    clock time of the visitor.
    Here is the breakdown for the site (this webserver is located in US central time zone so 0 is -6 GMT).  The drop is most telling on the weekend, affirming that readers prefer to spend their work hours reading on the Internet and their free hours doing other things.  I suspect for contributors it is the other way around, most of us spend free hours writing.
    Interesting Hank. But the server time is less informative, because it "smears" the information. In other words, this is convoluted with the distribution of visitors around the globe. If they all came from California you'd get a faithful distribution (integrating a distribution over a Dirac delta function returns the same distribution in output), but if they are unevenly distributed you get a hard-to-interpreted distribution.

    Can you post the reader time instead ? And in finer bins ?

    Interesting - however you should plot this in Universal Time, since people all over the world read your blog.

    For instance, I read your blog from the Bay Area, California, at my lunch time (~19:00 UTC)....

    Johannes Koelman
    "the fraction of night browsers, which is larger than I thought"
    Hmm... that could be people like me, who don't change the clock time on their laptops when traveling.
    Or people like me, who have homework to do but lack the will to do it.