Many companies are trying to understand and predict online social behavior using traditional marketing research techniques - both qualitative and quantitative. In most cases those companies are in for big disappointments.
Let's take a classic social phenomenon to make the point:
You have a crowd of 300 people who come to fill a theater with 500 seats. When they are all set and done, you will have clusters of people with big empty chunks of seats in between them, maybe nobody sitting in the front, and perhaps some surprising groupings of people.
How would a market researcher approach this situation?
The qualitative person would probably try to interview everyone ahead of time and make cluster predictions based on kinship, friendship, professional affiliation, etc. Not knowing when people arrive, nor understanding the true social motives for sitting in a particular place at a particular moment, most of those predictions would be wrong. Yes, maybe a couple will tell you that they intent to sit with a pair of neighbors, but when they get there, spot a potential client who they did not know would be there, and realize that the neighbors are not there yet, they might very well change their mind. There is no qualitative data, that you could have uncovered ahead of time, that would let you make that prediction.
The quantitative person would wait until everyone sits, lift up the curtain and take a snapshot of the sitting arrangement for further data analysis. The problem is that the data won't tell you anything. If there is a cluster of single women in the theater, you have no way to know, based on the data, whether those people were motivated by being with others who are just like them, or whether they maybe all came together as part of a mommy social group. If there is a cluster in the back, you have no clue whether those people were motivated by the desire to potentially leave early, or whether they wanted to be in a position where they could observe everyone else in the theater and just have a better people-watching vantage point. The data is meaningless when it comes to predicting social behavior.
So what can you do? You need to be more like an anthropologist and less like a market researcher. If you have the luxury to interview people ahead of time, and then watch the seating arrangement in progress, you will be able to make more informed assumptions, but you will still need to validate them through qualitative interviews afterwards. If you don't have the luxury of interviewing people ahead of time and see the seating arrangement in progress, you can still make assumptions and validate them through qualitative interviews.
But by focusing on understanding the parts of the whole through individual qualitative interviews or the whole by capturing data about the end result only, you will not learn anything meaningful about the true social drivers of this social gathering.
The lesson - don't try to understand online social behavior by doing traditional qualitative market research like interviews or focus groups (in which people will tell you what they want you to hear anyway), nor by doing sophisticated quantitative analytics research. Neither one will give you good results. Instead, focus on observing what happens, make assumptions and predictions based on basic human cultural behavior (need for status, need to hang out with like-minded people, need to impress others, being competitive among groups, etc.), and validate those assumptions through qualitative interviews and more observation.