Last time, I discussed testing the statistical difference in response rates. In this blog, we will tackle average gift size testing.
I’ll start with the problem we always run into: someone will give us the average gift sizes of a test and control, and ask if the difference is statistically significant.
The problem with that is you can’t use summary statistics for testing average gift sizes.
Here’s why: average gifts are susceptible to skewing. For example, you could have the control have an average gift size of $75 and the test at $25 and the two metrics still may not be statistically significant because the control may have had a single $5,000 gift that is skewing the results.
Therefore, in order to properly conduct statistical testing on averages, you must have the whole distribution of gifts. Each and every gift.
There are a number of different tests one can use to test significance. One of the most common is called a T-test. In SPSS, the stats software we at Analytical Ones use, this test is easy to run and analyze.
Just like in response rate testing, you also need to know the level of test confidence you are comfortable with. The level of confidence is indirectly proportional to the level of risk in making a change. Again, for direct marketing tests, we recommend using a 90% level of confidence. No one is going to die if we make a bad decision – unlike pharmaceutical testing.
The trick with all this is the need to look at BOTH average gift size AND response rate when evaluating the “winning” package. Often average gift size and response rates are inversely related. Meaning, as one goes up the other goes down. So, determining the objective of the test is important before declaring a winner.