Experiment design and analysis for e-commerce

Hi i'm new to the forum and hopefully somebody wants to offer me a bit of advice.

I'm carrying out a marketing choice experiment to estimate the effect sizes of copywriting new texts for our webshop. So to set up the experiement, i've made a webshop mockup with 4 similar products, one of which will get an four different options of desciption messages where 0 is the intiatial option and 1, 2, 3 are written by four different copywriters. As the subjects are divided into four separate groups by copy (0,1,2,3), so it's a between-subjects design for each experiment. I'm repeating this in four product categories, so in a way its either repeated measures or four separate experiments. Also, let me point out that as the respondent moves between experiments, he comes in contact with all four options: experiment 1 might test copy 0, exp2 might test copy 1 etc.

The problem is that i'm having issues analysing my own data which was produces by initial pilot testers. Basically i made a huge mistake in assuming i can analyze the choices with ANOVA but me dependant variable choice (0-3) doesn't meet any of the ANOVA requirements. I'm looking into binary logistic regression, but I'm confused as to how i should process the confusing dataset before analyzing it in SPSS.

So what i'm asking is whether I have made a massive error on my design or can this still be analyzed with meaningful and significant results on a sample of 500? How should the data be processed for binary logistic regression to measure the effects? Or can i still somehow use ANOVA? Any suggestions or comments on a better or simpler design are also highly valued. Data hasn't been gathered yet, but i'm a day away from sending out the experiment invites and have limited time to change the experiment design besides making some small changes.

Thank you for your help.
Hi again,

Sorry for the rather vague questions earlier on. I've successfully found answers to most of my questions already. Also seems I've successfully gathered the data and proceeding on with analysis.

Perhaps somebody can help me understand binary regression model fit tests. If I'm using a dependent variable as binary (purchase yes 1-yes/ 0-no), and an independent one (something was changed in the layout 1-yes/0-no) also binary. Can I still test goodness of fit with the Hosmer and Lemeshow Test? The test keeps giving me zero chi-square and calculate significance of -17976931348623157*(10 to the power of 292) in SPSS when I try to predict the outcome of the purchase (yes-no) with the independent variables. Omnibus Tests of Model Coefficients is however significant. I've checked with spearman rho and there clearly is a strong relationship between the variables.

What am I missing?