Assessing significance and incremental revenue from experiments

Occasional Contributor
Posts: 17

Assessing significance and incremental revenue from experiments

I am in the process of setting up an experiment by creating two groups (Treatment and Control). The treatment are new recruits who received special sales training, control are new recruits who did not receive this. The groups are independent and don't interact with each other. I review total sales from 2 groups 30 days post training. Each month new sales staff are hired and previous lot are asked to leave.

At the end of Month 1

Treatment: Size of Group: 50 individuals Total Sale: 750 Average: 750/50 = 15

Control: Size of Group: 30 individuals Total Sale: 350 Average: 350/30 = 11.67

I would like to test

a) is the difference in sales between treatment & control significantly different? (It has been a while I have done statistics, but this feels like a right-tailed directional t test, where we test if mean of treatment is significantly greater than mean of control?) - Is this correct?

b) if the difference is significant, can I say 3.33 (15-11.67 = 3.33) incremental? Is it right to say that £3.33 * 50 (size of treatment) = £166.5 is incremental sales revenue?

c) I have a total of 1500 new recruits over the next 11 months, can I say the total incremental benefit from this extra training is (1500 * £3.33) = £4995?

d) If the results are other way around, would I conclude this additional training resulted in revenue loss of £166.5? and thereby getting everyone to do this training will result in loss of £4995?

Super User
Posts: 11,796

Re: Assessing significance and incremental revenue from experiments

Since this sounds rather like class work I'm going to suggest some places to look.

Yes a t-test sounds like a good starting point. Proc TTest would be the procedure. Options to investigae would SIDES= to set the number of sides and direction. You can use the options TEST to specify a test for difference or ration  H0= to set a specific test value for the difference.

The data should be structured as: Group Sales for each person or a third variable to indicate the FREQuencyof identical values if you have lots of similarity and the data was provided in that form.

Discussion stats