January 10, 2011

Hashtag Analytics: Forecasting Engagement

Over the next few weeks, we're going to dig into some of the dynamics within the #measure web analytics community.

I retrieved data for an eight week period from early November to early January.  In the analysis, I looked at four weeks of history, seeing if there were trends that helped me understand who would participate again in the next week.

I created a segmentation scheme ... thirteen existing segments that describe user behavior. This will get a teeny-bit geeky for some of you, so feel free to fast forward to the results section of the discussion if you wish.
  • Recency = 1 Week, 1 Tweet Past Four Weeks.
  • Recency = 1 Week, 2+ Tweets Past Four Weeks, Participate Only 1-2 Weeks of Past 4.
  • Recency = 1 Week, Participated in 3 of Past 4 Weeks.
  • Recency = 1 Week, Participated in 4/4 Weeks, Skew = Broadcasting.
  • Recency = 1 Week, Participated in 4/4 Weeks, Skew = Neutral.
  • Recency = 1 Week, Participated in 4/4 Weeks, Skew = Being Re-Tweeted
  • Recency = 2 Weeks, 1 Tweet Past Four Weeks.
  • Recency = 2 Weeks, 2+ Tweets Past Four Weeks, Participate Only 1-2 Weeks of Past 4.
  • Recency = 2 Weeks, Participated in 3 of Past 4 Weeks.
  • Recency = 3 Weeks, 1 Tweet Past Four Weeks.
  • Recency = 3 Weeks, 2+ Tweets Past Four Weeks.
  • Recency = 4 Weeks, 1 Tweet Past Four Weeks.
  • Recency = 4 Weeks, 2+ Tweets Past Four Weeks.



Results:


The following table (click on the image to enlarge it) illustrates how the community "engaged" in a subsequent week.


Within this community, there are significant differences in subsequent engagement.
  • Those who just pop-off one tweet to broadcast something are not likely to come back.
  • The most engaged are obvious ... those who participated in four of the past four weeks.
  • Those who are frequently re-tweeted are the ones most likely to be engaged next week, a "duh", but interesting to see, nonetheless.
  • If a user goes more than two weeks without participating, it begins to become unlikely that the user will jump back in.
  • A new user only has a 10% chance of being re-tweeted.
  • Amazingly, new users made up 50% of those who were active in the following week.
  • New users, however, only yielded 21% of all tweets & re-tweets.
  • 25% of all activity (tweets + re-tweets) within the #measure community in this one week came from just 23 individuals (6%).  These folks participated in four of the prior four weeks, and skewed to individuals who are likely to be re-tweeted.

So What?


Yup, that's a common response from folks.  Well, here's what we're going to illustrate in the next two weeks.
  1. Next week, I will present a forecast model that shows me how the #measure community is likely to grow and thrive in the future.  I'll show how important it is to nurture the community, to encourage folks to participate.  In addition, I'll demonstrate the overwhelming importance of new participants.
  2. In two weeks, I will illustrate what happens to the forecasted trajectory of this community when one influential individual drops out of the community.
If you can forecast where your community is headed in the future, and you can understand the importance of new/valuable users, you can grow your community accordingly.
Interested in the forecasting aspect of Hillstrom's Hashtag Analytics?  Buy the book!