Usage Patterns for Cameraphone Driven Moblogs

Usage Patterns for Cameraphone Driven Moblogs

Eytan Adar

HP Labs

1. Introduction

Despite the significant penetration of cameraphones, there has been sparse publicly-available analysis of usage patterns. Usage may involve anything from the capture of images for personal use (picture stays on the camera or is uploaded to a user's PC), to person-to-person use (picture is e-mailed or messaged to another individual), to person-to-web use (an instance of moblogging). In this report we specifically address the last of these, and describe a simple experiment run in late January of 2004 on public data collected from a popular moblogging website
*. We find that while the median user takes 8 pictures in their first week (13.57 on average), the median number quickly drops to 1 picture a week within a month of use (4 on average), and a median of 0 by week 5. For reasons described below, we believe that these may also be overestimates.

Between August and December of 2003 an occasional crawl was done on the main moblog page for users that had posted new pictures. A list of 2780 such users was assembled. In January 2004, user blogs were crawled and postings (with dates) automatically extracted. During the full crawl we discovered that a large number of users no longer had an account. In total we were able to correctly crawl and extract information from 1448 users. For each user, pictures were placed into weekly buckets (bucket 0 representing the first full week that they user used the service, bucket 1 the second, and so on).

2. Results

Figure 1 depicts the distribution of pictures for each week until week 30. The boxplot illustrates the minimum (lower whisker), maximum (upper whisker), as well as the first and third quartiles (the box). We have also illustrated the median and average values which are barely visible in the figure due to the extreme maximum. Figure 2 is the same plot with the maximum whisker removed. Here the average/median behavior is much more obvious.

Figure 1

Figure 2

While we have plotted the usage patterns to week 30, we caution that data at the far end is noisy due to a small sampling size but is consistent with the general trend. Figure 3 shows the number of users in the sample that held accounts for different periods. While there are 1448 users that used the service for at least 1 week, the number drops off to 98 users by week 30 (i.e. ~7% of the users have had accounts for more than 30 weeks).

Figure 3

3. Discussion

It is very likely that the actual usage is in fact significantly lower than our data shows. This is for a number of reasons:

The results clearly point at a problem in the moblogging space. We have no reason to believe that there is anything flawed in this particular website's design relative to other solutions that lead to the behavior we see here.

We believe that the next step is to specifically understand why user interest in moblogging falls off dramatically. We can hypothesize that this may be related to design of the moblogging services, or to issues with the actual device. For the service providers such results should be indicate the need for better incentives for the posting of content. These incentives can be anything from a simple display of who and how many people are viewing images, the ability to rank images, better social networking, community formation, etc. More complex schemes would allow the images posted on the moblog to be pushed out to other medium. Integration into a user's blog or homepage is one existing example, but others are possible as well.

While incentives can come from the service providers, device manufacturers (and software vendors) have an opportunity to decrease the costs of using moblogging services. While bad resolution and image quality may diminish the desire to share an image, another major issue is the large number of key presses from the time a picture is taken to the posting. Users may be unwilling to deal with the hassle of the interface beyond the first few pictures.

While only preliminary, we believe that the results of this study are both a challenge and opportunity to service providers and device manufacturers who can properly incent users and eliminate usage barriers.

4. Acknowledgements

Thanks to Josh Tyler for early data collection. And as always, Bernardo Huberman for letting me explore new research spaces.

5. Further Readings

Last updated: May 9, 2004

*We have opted to anonymize the name of the service provider. We feel that the results are generally applicable and do not want to impact the business operations of any site in particular. We feel that the site is comparable to others in the space in size, users, and capabilities. top