Simpy Chichimichi

In the Timing of Read vs. Write study we saw that there is a slight difference in when Simpy users Write (save and tag new links) and when they Read (search for previously saved and tagged links). In addition to this difference in timing, I've been monitoring Simpy's web access logs and have always found it interesting how its access pattern differs so much from web access patterns I was used to seeing on other sites. This particular study is not about tagging per se, but I kept "Tagging" in the title for the sake of consistency.

First, let's look at the access pattern of a simple, static web site. In this case, the data belongs to Lucene Consulting site. The following image shows the self-explanatory Weekday-Hour map, where a lighter colour indicates a higher number of web page views.

Clearly, the number of web accesses is pretty evenly distributed throughout the day and throughout the week, with Saturday and Sunday bringing in less visitors. I think this is a pattern you will see on most web sites. On sites targeted at the business crowd, or any audience that uses those sites for work, the curve between 09:00 and 17:00 will be much more pronounced, and will look a bit like a bell curve. Also notice a slightly higher number of visits around 22-23^h (the time is in EDT, +0600). Here is another chart showing the same data in a more conventional format:

How about a blog site? What usage patterns can be seen there? For that, let's look at data for this very blog:

Same data in an alternative chart:

What do we see? We see a very different usage pattern. There is no 09:00-17:00 EDT curve whatsoever, there is no 22:00-23:00 EDT bump we saw on the previous set of charts, but there is a major spike between 00:00 and 01:00 AM, and there is a significantly increased nocturnal activity after 00:00 AM EDT, until around 04:00-05:00 AM EDT. What could be the cause of this? From what I have observed in the logs, this activity comes from FeedFetcher-Google, Onfolio, Bloglines, NewsGator, Rojo, kinjabot, other blog aggregators that, like bats, seem to come alive and start feeding at night. In my experience, this pattern is not limited to blog sites. I see the exact same pattern with hyperactive nocturnal bots in Simpy itself. It's no wonder, there are lots of pages to be fetched from Simpy, and there are thousands of RSS, RDF, and Atom feeds to be fetched. I wish major blog aggregators published this type of data, because what I would really like to see is the human usage pattern, not that of bots. Anyone?

This is really becoming a pattern:

Lots of initial buzz, followed by a quick drop in numbers back to previous levels. Interesting...

About two weeks ago we analyzed tagging in the context of public vs. private links and concluded that Simpy users put more effort into tagging when they tag their private links.

I tend to like observing macro-level patterns of human behaviour. For instance, observing different patterns of groups of people due to their different cultural background, gender, age, etc. I also like observing changes in behaviour over time, and trying to correlate those changes to various events. This interest of mine often leads me to the BlogPulse Trend Search. Combine that with my love for charts, and this leads us to the topic of part 3 of this Web 2.0 Tagging series - time.

Time is an important factor in a number of Web 2.0-type services, most notably in the world of blogs, where it is commonly used as the default sorting criterion. Some people refer to Web 2.0 as Read/Write Web, and I was curious about when this Read and Write activities happen. Do they follow the same pattern? Do we Read as much as we Write, or do we just save stuff and never go back to reading it (@toread, anyone?)? Do we Write more in the morning, and if so, does that indicate that people like to gather new information in the morning? Does that follow the "read email first thing when I get to work" pattern?

I'd like to be able to ask all these questions and more, and get answers for them, so I can have a better understanding of us humans. While, unfortunately, I don't quite have all the detailed data that would be needed to answer these questions properly, I will share what I have. The following chart shows the Write activity - it shows the number of link additions broken down by the hour of the day. The stacked graph is the result of me taking two snapshots of the Write activity, some amount of time apart. Please note that this uses the +0600 timezone (EDT).

What does this chart tell me? It tells me that Write activity is spread throughout the day pretty evenly. However, this chart represents the aggregate activity! What would be much more interesting to see would be Write activity from an individual user's "time-of-day point of view". Reminder: please update your info and enter your city or postal code, or at least your state or country, so I can provide your with various geolocation-related services for which I have ideas, but need data.

Because this information is at the aggregate level, and knowing that the timestamps are from the EDT (+0600) timezone, I can draw the following conclusions:

• Europeans are not very active, at least not during their morning hours (3-5 AM EDT on the chart). However, this doesn't quite match the demographic data people have provided, which puts 6 Western European countries in top 10 countries Simpy users are from. If you want to represent your country, please update your info.
• The east coast in the USA starts being active around 8 AM EDT. Don't count me in - I'm still asleep at 8 AM!
• The increased activity around 13^h (10 AM on the west coast in the USA) might correspond to the west coast people waking up, getting to work, and doing their bookmarking in the morning.
• The peak around 22^h is a bit puzzling. Perhaps this comes from people in east Asia (22:00 EDT is 10:00 in the Far East), or perhaps this comes from evening/night activity of users in the U.S. (you can count me in here).

Next, let's look at the Read activity by looking at the chart that shows the aggregate number of searches over the course of a day. Again, two snapshots were taken on two different days, and the times are in EDT (+0600) timezone.

This chart again points to lower levels activity among Europeans and possibly Asians, as serious action starts only at 7 AM EDT (12:00/13:00 in Europe, and 19:00 in the Far East). Again, action seems to start when east coast, USA wakes up, and again there is a little peak around 13:00, which corresponds to 10:00 on the west coast. In the evening we again have a slight increase in activity around 22:00 EDT (10:00 in the Far East).

Overall, it looks like Read/Search activity exhibits a much bigger "hump", while Write/Additions are spread more evenly throughout the day.