How many customers? 10s 100s 1000s?
It would be a bit tough to pair samples, since you are looking on different days.
Hi guys, hope you all had a good weekend.
I am wanting to analyse the effectiveness of a promotion on the amount of money that was spent in shops by specific customers every day.
If I have on a daily basis, say, a customer's turnover per day before the promotion started - and the daily turnover of the same customer per day after the promotion started - what is the best statistical method I can use to determine whether the promotion had a significant uplift on the turnover of a customer?
I am thinking a t-test on an unpaired sample set but I wasn't sure if this was the best way.
Any help would be much appreciated.
I have the possibility of comparing 1 customer up to 1000 customers.
I agree about tough to pair samples. I am thinking I can compare the 30 days prior as 30 values in an unrelated set, to the 30 days after as another 30 values in an unrelated set for an individual customer and see whether the second set is significantly different (higher, really) than the first set.
Or, for each of my customers rather than 30 values in a set, I can have 2 values - the total turnover in the 30 days prior to promotion and the total turnover in the 30 days after promotion. Those could be paired and then I have pairs for my x number of customers, which I can then analyse to see if for the group there is significant difference.
You must be very careful not to miss things that mustn't be missed. For example, if your first 30 days are right beore Christmas and the second 30 are right after, you WILL get the wrong impression concerning your promotion. That one is obvious, but less apparent ones will mess with your head.