Here is an awesome summary of some of the discussions at GDC about the uses and pit falls of user analytics.

To me one of the most interesting issues is trying to understand user engagement via analytics. I can see there is huge scope for important UX work here but in terms of emotional engagement I think the data is near useless. This data is collected primarily for UX and commercial considerations and its a terrible fit trying to measure narrative or emotional engagement and immersion; the figures are skewed heavily by different kinds of play and even in data from a single player, by different moods and situations. It is very possible that when concentrating fully on the gameplay the player may find themselves deeply immersed in the story. Many mitigating factors that cannot be identified easily by analytics can interfere with the player’s concentration which may not paint a useful picture of overall engagement. I can see that some kind of averaging of player behaviour over time might give a vague indication of player engagement but aggregate data from a large number of players is useless in this respect.

Perhaps the wrong data is being collected, but I am not sure which would be the right for this specific purpose…