Video #3 [Open Playtest]
They read all the comments.
The surveys were extremely important to the final product. Absolutely essential.
They figured out out to best iterate with a purpose on problems later in the process and wish they had figured that out earlier in the process.
People's ratings of the game went up and up with each packet of the playtest. There were minimum quantity thresholds for feedback before data could be considered sufficient to consider.
The forums are not necessarily representative of the larger audience. There definitely is a silent majority sometimes. A lot of times people would say something is terrible on forums, that came back with 95% approval on surveys. It was most useful to use forums when the forum views lined up with the survey data, where they could then ask forum people more about that aspect of the game. Also sometimes the thing people would complain about was more a sign something was going wrong in a broader issue, and not what they would specifically complain about. Like for example someone might complain about not hitting enough bad guys, but they'd find that was really a symptom of a cause of lack of sufficient movement for PCs.
Here is an image of a sample of some of their data, measured on a 1-5 scale of satisfation. NOTE: They specify the headers are off on this.all three columns are off (if you can even read them in this grainy image - supposed to be combat satisfaction, non combat satisfaction, and average satisfaction overall, but numbers may not be in that order). This was not the headers they worked with he just knew they got moved before the presentation by accident:
Here is another on Class Complexity from a prior set of data:
There was another chart here on Complexity vs Non-Combat satisfaction and they found classes high on non-combat satisfaction tended to also be high on complexity. So for non-combat, people liked the complex classes.
The reverse was true for combat. Audiences said they were having the least amount of fun with the more complex classes in combat, and the most amount of fun with the least complex classes in combat.
This set of conclusions was the complete reverse of the designer initial assumptions about the game. The assumption was that during combat people liked having a lot of options, and that during non-combat they wanted a lot more free form and not as many mechanical options. It turned out it seemed to be a time pressure issue. When out of combat, planning stuff was more fun with lots of options to work with and time to flip through books and such. But once in combat and everyone is waiting for you to complete your turn you don't want to have too many options and flipping through books resulting in slowing the game down.
Here is an image of Rogue Satisfaction from October 2012. I am sorry the image is so grainy I couldn't seem to grab a clearer one.
Cunning Action was being rated at 100% (40.1% apparently rated it top, and it had the top rating of being above everything else) as the top rated element of the rogue. Cunning Action was at the time of this survey part of the Thief subclass and not part of the core Rogue abilities. Thieves Cant was getting the lowest rating at 3.1% (which they didn't care much about because they viewed it as a ribbon mechanic.)
They followed this up with a grainy image of two different Rogue surveys (August and October) charted against each other showing where there were response improvements from changed in the playtest for that class, and still some dips that fell below their threshold for acceptable that continued to need work.