Ordinal data analysis and thesis testing

Hello,

I have been working on a survey analysis and have to test hypothesis if some kind of software resulted in increased performance of a company. The data are collected using questionnaire where the respondents had to choose the right answer based on Linkert ordinal scale:

-1 for making the performance even worse
0 for remained the same
1 for slightly increased
2 for significantly increased
3 for extraordinary increased

I interviewed about 40 respondents.

There are 50 questions and they can be grouped in 4 performance categories that I have to test, and also I should have overall test.

So, the central hypothesis is if the software is making positive impact on the performances as whole, and there are 4 auxiliary hypothesis related to those 4 groups of performances.

So, now, I have results in the table (excel, spss), and have some questions:

1. do I have to group those answers before the analysis (e.g. to sum the results in those 4 groups) or I can work with the whole table (meaning 50 columns)?
2. since I used Linkert scale, the data are ordinal, so I suppose I shouldn't use mean to make average value for groups?
3. what kind of hypothesis testing method should I use (ordinal data, many columns-50, but the same respondents), and should I first group the data and perform the test?

Thanks in advance
DD