I want to classify retail customers into groups reflecting the periodicity of their purchasing behaviors: e.g. weekly, monthly, seasonal, annual, etc. I thought this question was trivial, but now find myself stumped. Initially I attempted mean/median days between purchase but found this did a horrible job classifying "cluster" purchasers - for example, a customer for whom visual inspection reveals an annual/seasonal buying pattern based on the following purchase dates but mean/median fails:
- 9/5/18
- 9/11/18
- 9/18/18
- 8/29/19
- 9/5/19
My goal is simply for a nudge in the correct direction from the community. I'm certain this has been solved in other domains (or perhaps even this one!) and I am just struggling with my own ignorance. Is this type of modeling/analysis addressed via time series? Signal processing? Thank you in advance for any insights.