## August 16, 2020

### A Bit Of Geeky Math

Here are the results of a Logistic Regression used to predict how likely customers are to purchase for a second time within twelve months.

Don't worry about what all the geeky numbers have to say ... I didn't even bother to give you column headings.

A few findings.

• Customers acquired in January / March / May / June / July are more valuable than customers acquired in other months.
• Customers acquired in December are about 10.4% less likely to repurchase than are customers acquired in other months. What month do you really crank up the digital advertising to generate new buyers?
• There are merchandise categories that deliver customers with better probabilities of repurchasing ... Categories 03 / 04 / 05 / 07 / 09 yield customers more likely to repurchase.
• "AD1" in this case represented new customers via print. They were 63% more likely to repurchase in the next year than those acquired via other advertising channels. Hopefully the brand didn't spend 75% more to generate an increase in repurchase rates of 63% (hint - that's what frequently happens).
• Customers who purchased because of discounts/promotions (kev_disc) were 3.4% less likely to repurchase in the future.
• Customers who bought items (ABO) that sold for above their historical average (i.e. paying \$34.95 for an items that sold for a historical average of \$32.95).
• Each item purchased, all things being equal, adds 5.5% to your first-year repurchase rate.
When a customer purchases for the first time, I'll bet you immediately score that customer and then calibrate your Welcome Program accordingly, right? In other words, if the customer is in the bottom 20% of all first-time buyers, you treat the first-time buyer differently than if the customer is in the top 20% of all first-time buyers. Right?