I concur with this. I think people can give people higher up in corporations too much credit for how informed they are. Some people in higher positions can be pretty clueless and data can be bad. I can say from experience where I have worked for such a company.
I don't think the dev team is as bad as what I had, but I think it is plausible that the dev team could base decisions on bad data.
For instance, if 90% of people who played healers in their original design were happy with it. What statistics would show this? Very few people post anywhere that they are satisfied, they just play it. I mean, it's taken me until 5.0 to give any feedback on healer play and I've played healer since BETA and was a satisfied healer during most of that time since.
Then take those who're dissatisfied. Take, for example, people who keep forgetting about Cleric Stance or people in parties with healers who forget Cleric Stance and wipe. People then complaining about cleric stance, it's too clunky, it causes wipes, it leaves room for easy mistakes and so on. And if this is all you see about Cleric Stance, then it's plausible the data would suggest "Cleric Stance overwhelmingly bad" and then remove it.
But there are many of us who liked it, who thought whilst yes it could be improved but it should not be removed. This voice didn't come out until after Cleric Stance was removed. Again should people complain their healers DPS too much or are making too many mistakes with their DPS focus, without people saying "I like how this works" it looks more likely the right thing to do is simplify how they DPS.
And chances are, if say, they were to look at the forums now for people's satisfaction, what they're going to see is the complaints about those who're dissatisfied, rather than what's working and why for those who are satisfied.
I hope their means of data gathering is good, however, I'm not sure that it is. But I'd see no shame in them putting out surveys to gather such data before making decision. I think it'd be a more transparent approach and if the survey is well constructed then people could have the opportunity to provide more meaningful feedback than say, our anecdotal rants on the forum. Whilst anecdotal rants on a forum are great and putting things into context for certain complaints, it's not good for statistical data about what people like/don't like.