Friday, September 25, 2009

Twitter Usage Remains a Bit Mysterious


Sometimes numbers can be deceiving. Hitwise data, for example, suggests that Twitter traffic hit a peak in April, and then dropped, peaking again in July before dropping to levels below that of the April peak. In other words, traffic has dropped since April.

But one has to make adjustments. A majority of Twitter users use a third-party client to access Twitter. In fact, only about 20 to 30 percent of people go through the Twitter Web site. So the direct Twitter data does not show the full impact of Twitter usage.

The Hitwise data also suggests the same effect, showing the number of new users--new, not total--coming to Twitter from its top traffic sources, such as Facebook, Google, and MySpace, also has fallen consistently across the board from April to September 2009.

But those statistics only point to a slowing rate of growth, and that always happens to any application or service which starts from a low base, no matter how popular.

The Twitter Web site does attract about 54 million visitors a month, and that does not count the 70 to 80 percent of users who use the application from a third party portal of some sort.

A slowing rate of growth likely is nothing to worry about. Other studies have shown a high abandonment rate for new users, though.

In March 2009, for example, more than 60 percent of Twitter users fail to return the following month, says David Martin, Nielsen Online VP.

That means Twitter’s audience retention rate, the percentage of a given month’s users who come back the following month, is about 40 percent.

To put that in perspective, it is roughly the equivalent of turning over 100 percent of the user base every three months. Such a churn rate is unsustainable. One suspects the churn rate also will drop over time. People either like it, or they don't. But whether they like it or not, even resisters will use the app if their friends, family and associates do. Ultimately, that is going to make a difference.

No comments:

Costs of Creating Machine Learning Models is Up Sharply

With the caveat that we must be careful about making linear extrapolations into the future, training costs of state-of-the-art AI models hav...