Structure your data as attached.

You can actually achieve the same structure as far as Tableau is concerned without physically changing your source data, by using a custom SQL connection.  That would get quite messy if you have lots of questions, though.

I've attached a copy of your original workbook with a custom SQL connection to your original workbook showing how to do that.  The custom SQL basically just picks each question out from each row and turns them into separate rows.  It looks like this:

SELECT ['Survey 1$'].[Location Code] AS [Location Code],

  'Question 1' AS [Question],

  ['Survey 1$'].[Question 1] AS [Answer]

FROM ['Survey 1$']

UNION ALL

SELECT ['Survey 1$'].[Location Code] AS [Location Code],

  'Question 2' AS [Question],

  ['Survey 1$'].[Question 2] AS [Answer]

FROM ['Survey 1$']

UNION ALL

SELECT ['Survey 1$'].[Location Code] AS [Location Code],

  'Question 3' AS [Question],

  ['Survey 1$'].[Question 3] AS [Answer]

FROM ['Survey 1$']

UNION ALL

SELECT ['Survey 1$'].[Location Code] AS [Location Code],

  'Question 4' AS [Question],

  ['Survey 1$'].[Question 4] AS [Answer]

FROM ['Survey 1$']

The custom SQL would start getting quite long if your real data has lots of questions (i.e. is very wide) - I know you can run into a limit in the MS JET database engine used for connecting to spreadsheets if you have too many clauses to UNION together.  I vaguely remember someone posting to say they had hit the limit at around 70 or 80 - but then posting back to say that they had found a workaround which involved breaking it into chunks and nesting them.

It really depends how much pain physically restructuring your original data source would cause.

I presume you mean columns where nobody chose that answer, do you?

The simplest thing you can do is un-check the Hide Empty Columns option under the Table menu.  That will display all columns (i.e. all values which exist for any question for any location) even when you have filters set which mean there aren't any entries for that column.

Bear in mind that Tableau can only show columns (answers) that it knows about - so if there was a possible answer which nobody chose at any location, Tableau won't know about it and can't show it.  If you think about it, there's an infinite set of possible answers which nobody chose - "Who Cares" being the typical one that I always want to answer to all the questions on nearly every survey that gets stuck under my nose.  ;-)

That will give you the columns, but it won't display 0% - because there actually isn't a value there at all.  If you really, really want to display 0% that's harder - it almost certainly means you have to mess with the structure of your data source again.  Post back if you do want that - and we'll see what Joe can come up with.

If you are interested in making a visual display of survey data like this, check out the resources at http://www.datarevelations.com/category/vizbizwiz-blog

Yep, you can do that.  I couldn't think of a way to do it with filters, but you can achieve almost the same with a parameter and a calculated field.  The only drawback with that approach is that the parameter won't automatically pick up any new Centres that get added to your data.

I added a second column with the overall answer.  I also added a graphical view of the same thing in the second sheet.  By locking the axis scale it makes it quite nice to spin through the Centres and see how the responses change.  Finally I added a view which compares all Centres with the overall score.

Thought about this again while I was out for a run.  I've added another view which does let you filter the Centres whilst still keeping the overall comparison, so you avoid the issue of the fixed parameter value list.  I've done it as a multi-select filter so you can compare one or more Centres to the overall results.

Attached is what I was thinking of for a good view of this kind of data.