A Data Science Central Community
Most of grocery chains offer club member cards to customers so that customers can get discount when shopping in stores. I would like to know how to measure the effect of such programs. What directions should I take?
If you don't have any suggestion to this question, I would appreciate if you can point out good books covering this subject or other discussion boards/LinkedIn groups where I can post this question.
Simply collect sales data for a specific period from stores with and without club membership cards. Run an ANOVA to see if there is any significant difference between the means for the two sets of data. Hope it helps...Ravi Sangal
This would be a project that has a lot of facets to look for in the data. Businesses want to know not just the numbers but what the numbers mean and how the numbers can be applied. It is not an exact science but this allows you to get closer to what is to be expected for sales and promotional.
If the establishment is big enough to offer and track customer cards, you will be able to do things like elasticity, compensated demand analysis, profit maximizing, cost-volume analysis, consumer preferences, advertising dollar effectiveness, consumer targeting, purchasing power correlated with the macroeconomic income information, and statistics to determine the lag effect with consumer response to the new advertisement. And the list goes on. Some of these you should do any way but when volume is expected to change, they become even that much more important.
Even if the customer cards have no name on them, you can still do a lot of the above. This will allow to go across transactions even though they pay cash. The credit card/debit card and checks have the same type of functional use but not everyone uses plastic to pay. Businesses are allowed to store CC main numbers but on the credit card they cannot store the swipe information.
There is a lot that can be done.
Let's make the scenario simpler. Suppose a company selling some niche products to business customers. All business customers have an account with the company so that you know when a customer closes its account. Also all transactions are recorded. Some customers are offered a particular discount program. I want to see if this discount program helps customer retention. How to go about this analysis? Here is my idea: select two groups of customers. One has customers which were current and on the program 2 years ago(case group). The other has customers which were current and not on the program 2 years ago(control group). Then use survival analysis to see if case group has higher retention rate than control group in last 2 years. Is this a valid way to conduct this analysis? What are the alternatives?
The niche products to business customers scenario is quite different from grocery and club members. With the niche products, you will definitely be required to know your market to do what you are asking. Niche products requires you to know the market demand, competition, and product lifetime(i.e. an ice cream cone is very short and the customer comes back....a desk has a very long lifetime and the customer is highly unlikely to come back the next day).
Now to answer your question, I am of the opinion that the measurement for customer retention is volume of sales($ and quantities over time) and frequency of transactions(count the days between transactions over time). So if any of those two decrease consistently over time then the customer retention is going down the tubes. Also if the customer has a net 30 terms, check the consistency of payments and the only thing to worry about is being late for no reason. This is the simplest checks to do and to model.
These are inputs into your scenarios of have the customers on the program and the customers that are not on the program. Then you will be able to find out a good discount program for your customers.