In a lot of my projects in 2019 there is a "lurking merchandise problem". In other words, it's a problem that isn't easy to identify on the surface.
But a simple modeling technique identifies problems.
Take all customers who purchased 13-24 months ago, and for that year sum up their annual spend in the 13-24 month timeframe by merchandise category.
Then create a dependent variable for total company spend in the past 0-12 months.
Regress 13-24 month variables against 0-12 month spend (after cleaning up outliers, of course).
Merchandise categories with smaller-than-average coefficents are categories that dissuade customers to repurchase.
Merchandise categories with larger-than-average coefficents are categories that encourage customers to repurchase.
This is an old technique ... we ran this stuff at Lands' End in the early 1990s (yes, I said 1990s ... early 1990s) and learned that Home purchases dissuaded subsequent orders (because the customer needed Home items less often than the customer needed a mock turtleneck).
It's common to find a merchandising team that expands into a category that dissuades future customer loyalty. And your merchandising team probably doesn't have an analyst running regressions against the customer file, so how could they possibly know that their decision was a bad one?
If you don't have the resources to do important work like this, give me a holler (email@example.com) and I'll do it for you.