Before we get too far down the technology path with our Convenience Store 2.0 project, let’s decide what we want to do.
And before describing what we want to do (increase turnover, but how?), let’s figure out what we’ve got first.
Just Give Me Your Data
This convenience store chain has a network connecting roughly 20 stores in 2 states. Each transaction is sent back to the ranch to be stored in their database. Inside of the database, we have tables for:
- Companies (they own more than one brand)
- Regions
- Stores
- Products
- Transaction Headers
- Transaction Details
- Costs (calculating cost in this environment isn’t as easy as one might think)
- Prices
- Employees
This collection of tables roughly represents what I call the ‘Operational Data Store’ or their Online Transaction Processing region. Missing from this, and what we’ll add in the ‘Dimensional Data Mart’ region, are tables that describe additional dimensions for analysis, like time (this one is important and will become even more important when we bring this whole thing into a real-time analytics mode).
And The Questions Are?
We’ve got a lot of questions, but we’re going to start with something very simple. But first, our goal again is to increase turnover without cannibalizing and to increase turnover in products with higher margin. This is important – this simple requirement will end up dictating how much money we can spend on the solution given the expected results (and we’ll model that too).
So, here are some initial and simple questions that we’d like answers to:
- When someone pays for gas inside (discount), what items do they buy?
- When someone buys a particular item, what other items might they buy?
- How does this differ by region, city, state?
- How does this differ by time of day, day of week, etc?
This will begin to give us some insight into the spending patterns of the customers. And might lead to some very interesting decisions – like should we have meta-regions based upon buying habits and not just geographic location? Can we prune items from inventory shifting capital to more productive products or product lines?
Also, this work will begin to lay the foundation for our real time marketing campaigns. Location based stuff isn’t going to work (unless we implant RFID tags in their customers), so we’ll have to go on a store by store basis. The end goal (increase turnover in higher margin products) will be supported with a large flat panel screen in the entry way of each store (one or two to test) that will flash specials. We need to make sure that the specials are working, kill them if they’re not, and let them run longer if they are.