Dear board members,

I have conducted a field experiment using Facebook ads. Now that the experiment has nearly come to its end, I have started to think about what tests are adequate and how to do them.

Basically, I used a Difference-in-Difference approach (2 groups, 2 periods) to measure whether Facebook ad A performed better concerning Click-Through-Rate and Conversion Rate than Facebook ad B. However, now I have the aggregated data (sum of impressions e.g. 60,000, sum of clicks e.g. 430, ...) for each advertisement and period. Yet I have only used SPSS and only for testing data from questionnaires, i.e., non-aggregated data. Therefore, the aggregated form of the data provided by the experiment poses a challenge for me, since I do not know how to test whether the ads significantly differ in terms of Click-Through-Rate (i.e., average clicks per impression) and Conversion Rate (i.e., average converison per impression).

I thought of doing a Mann-Whitney U-Test, but I am unsure whether (a) this is the right test and (b) how to do it properly in R or SPSS.

I hope you can help me.


Greetings
Simon