A Simple Survey -- Critique Me!

#1
Hi everyone! First time poster here! :wave:

I'm designing a survey for my small business -- but unfortunately, I know very little about statistics. :( I have a survey in mind, but I'm wondering if the methodology is flawed. So if you're knowledgeable about statistical research, please critique this project to ensure that my results aren't skewed! :)

So let's say I sell cookies. And I want to see which offer is more enticing for customers -- If I sell them for 50% off, or if I do a "Buy one, get another for free" deal. (By the way, this is just an example [I don't actually sell cookies], so just humor me in thinking that a survey is the proper venue to explore this.]).

So I conduct a survey. The first question is, "How likely are you to purchase my cookies in the near future -- Very likely, likely, unsure, unlikely, or very unlikely?" Then after some filler questions, I tell them the deal I'm testing -- "Next week, I'm selling my cookies for 50% off." Then I ask the first question again, except re-phrased: "On a scale of 1-5, how likely is it that you'll purchase my cookies in the near future?" And I see if there's any change in the way they answered the first question and the re-phrased one (as an indication of the deal's effect).

After this, I'll be conducting another survey. It'll be exactly the same, except the deal that's between the sandwiched key questions -- This time, it'll be "Next week, if you buy a cookie, you'll get another for free." And again, I'll be comparing the responses of the first question and the re-phrased one.

So after all all this, how should I measure the overall effect? I was thinking of just subtracting the differences between the two key questions (i.e. Let's say one answered "Unsure' at first [which is a 3 on the 1-5 scale], then answered 2 [which is "Somewhat likely"] on the re-phrased question, coming up with a +1 change) and averaged all these differences into a single mean -- and compared these two means to see which deal is more effective. Is that an adequate methodology?

Or should I just use a percentage? Let's say 60% of the respondents increased their likelihood after hearing the deal in the first survey. What kind of assumptions can I make from this?

Anyway, I have a million questions, but I'll just leave it at that for now. Again, this is just an illustrative example -- please only critique the methodology of this survey.

Thank you so much for any input you can give! :)