Beer and Nappies - Data Mining Urban Legend
Interesting post over at blogs.sas.com pointing to an article by Bob Garfield that dispels as urban legend the Data Mining anecdote that putting beer and nappies together increases beer sales due to the fact that Dad’s pop in to buy nappies on the way home from work and pick up some beer as well.
Must admit I have used that line a few times in the past, but it does highlight the idea that Data Mining can identify differences or similarities that would not normally be pciked up via simple data exploraton techniques
A while ago I was luck enough to be involved in a Data Mining exercise for a large supermarket chain, and a few things came out of the analysis that we probably wouldn’t have picked up manually:
- One of the stores had a high proportion of shopping carts with meat in them, the other was primarily vegetarian
- Banana’s were the major linkage product for all carts in all stores
- There was a certain customer segment that regularly brought cat food, chocolate and red wine
(I still think naming market segments is a fun task!) - House brands vs non house brands were dependent on a certain percentage price difference
- People would regularly buy house brands and non house brands in the same cart
So would changing the layout of the store based on some of these preferences increase sales?
Good question…
All I know is that this insight took a few days of effort, it probably would have taken months to do it manually (and that is assuming we would find these results).
Thats the power of Data Mining!