Do you take a top-down or bottom-up approach when you do your post-campaign analysis? Do you look at households, individuals, items, orders or recency?
Your answers will tell a lot about your data processing, and what you can expect from your planning.
Merchandise can have short or long ordering windows. A clothing purchase is usually an all-in-one order. Shoppers know that if they wait to get the sport coat that matches the slacks they like, their size may be out of stock.
Electronic products like stereos, home theaters, and computer systems have much longer windows. A student might buy a notebook at home and wait to buy the docking station, extra keyboard and monitor until they are at college. Audio and video hobbyists add to their systems over several days, weeks or months.
Like most mailers, you probably have stuffed your database with mailing treatment keycodes, household and/or individual ID numbers, names, addresses, shipping addresses, order and shipping dates, units, SKUs, cost and retail dollars, order numbers and line numbers.
But have you really looked at all this in a holistic way?
Let’s start with an apparel purchase. Several line items ordered on one day can be summed to a single order number and purchase dollar total, giving you one address and ID number.
You can add up the line items on this customer’s record by ascending or descending date, by state or ID number—you will end up with the same information in your summary. Your modeling and geography will not change.
Second, imagine the student who buys a notebook computer that she plans to load with games and music at home. Several days later, she orders accessories to be delivered to her college in a different state.
Your system records several line items under different order numbers that share the same ID number. Typically, you will sort the file by date. The date and address on the last record will be the one you keep.
But here’s the rub: Your RFM data will be accurate, but your geographic summary will put all the dollars in the same state as the college. If you know that split orders such as this often appear in your data and sort your file to capture the correct address, you will keep the first order date and your recency of the buyer’s activity will be incorrect.
This is a very picky view of how customer transactions can be handled. But you may not understand your customer’s purchasing habits until you align your analyses with real behavior.
Bill Singleton writes “Show Me The Data” each month for the Lists and Data Strategiessection in Multichannel Merchant.. He is a manager of analytics and consulting services at The Allant Group in Naperville, IL. He can be reached at: firstname.lastname@example.org and 630-579-3448.