[RFC] Sharing of survey results and impact on questions asked

[Ian Clatworthy]

All,

I need some advice real soon on what I ought to be sharing in terms of
registration survey results, how often and what that means in terms of
questions asked & promises made. Privacy of information is, rightly so,
a very sensitive topic so we really need to get this right before the
survey goes live next week.

Bazaar is a community project so I feel that the community owns the
data, not Canonical and certainly not me. But sharing everything,
including IP and email addresses, isn't necessarily the right thing to
do either, *particularly* if we're giving users the option of saying
whether their responses can be shared or not.

My initial thoughts were that I'd put together a report once every few
months and forward it to the list. But "self-service" is a viable
technical option so, being lazy and looking to avoid unnecessary work
for me, that seems a far better direction to go. Here's the web page
sharing the (recent) results.

http://www.surveymonkey.com/sr.aspx?sm=FDeDsLzaZ0AKNEATuA0QklWayMZKGfHcrP1U4V55jAY%3d

I made lots of changes yesterday afternoon following the feedback I had
received until then so I had to clear the data received before then.
Even so, there's enough there I hope to give you a feel for how it will
work. Note that the move to named features (away from text boxes) works
*much* better. I downloaded the original data (26 responses) into a
spreadsheet so I could use those initial responses to seed feature
names, etc. and the updated survey is much easier to get actionable
information out of.

Now, there are multiple levels of sharing of data permitted. See
http://www.surveymonkey.com/HelpCenter/Answer.aspx?HelpID=30. It's
currently set to the maximum amount and I can easily take things away if
appropriate, e.g.:

* ability to download all data
* ability to filter
* ability to browse individual responses
* ability to see textual answers

Even with all those things gone, the summary is quite useful, so that's
one easy option that hides individual responses within "bulk
statistics". What I can't seem to do is easily filter answers based on
multiple responses to a question, e.g. the current Q re sharing isn't
viable IMO.

So here are the questions I need answers to:

1. What level or access to the data is right?
2. What do we commit to accordingly?
3. What questions do we remove altogether accordingly?

As a final comment, keep in mind that all comments to the mailing list
and on Launchpad's bug tracker are public. So I don't think we need to
hide data, *unless* we commit to doing so.

Ian C.