A novice in deep waters

Hi guys

I am in no way a statistical expert – my knowledge about the subject is limited to a few courses I participated in a couple of years ago at the university where I studied.

Therefore the question below is probably far from the level on which you typically discuss statistics.

I am employed as a project manager in the Telco business where I am in charge of a project which aims to strengthen our customer satisfaction when in contact with our customer service agents.

At the moment I am redesigning the way we evaluate our customer service agent’s transactions with our customers whether that being on phone, chat, or email.

My goal is to identify what we call Customer Critical Errors, which is thing one must not do when in contact with a customer as this will affect the customer satisfaction.

At my disposal I have approximately 230.000 surveys from the last three years that are based on transaction with our customer service.

The surveys are formed as:
1) How friendly was the agent?
2) To what degree did the agent do an effort in solving your problem or answering your question
3) To what degree did you think that you got a thorough and competent answer
4) How clear was the information you got from the agent
5) How long did you experience the waiting before you got through to the agent
6) Was your problem solved?
7) Based on this contact with our customer service would you then recommend our company to family and friends?

The first five questions are on a 5 point scale (0 = not at all’ish, 4 = very much’ish)

Question six is a yes/no

The last question is a 10 point scale (again; not at all/very much)

This last question is also the one we use to assess the overall satisfaction.

My thought was to do regression analysis in order to identify which of the first four questions have the biggest impact on the last question (since those four are the ones which the agents themselves have control of)

But then again I am thinking that I cannot just ignore question five and six since I assume they always have an influence on the answer I question 7.

So what should my approach be? Is the regression analysis the right tool to go with at all?

Also I assume there are some methodical pitfalls since the first five questions is on a 5 point scale, with the sixth being a yes/no and the last a 10 point scale?

Our budget is limited so I am not able to just outsource this to a statistical firm and I am not seeking a "free" analysis.

I know this is a “big” and perhaps not simple at all question, but if anyone have some inputs/directions then it would be much appreciated.

Thanks in advance

Best Regards