Measuring Customers

Aggregating at Different Time Periods

• Typically, it is useful to measure features for different time periods
• By doing this, we can integrate a measure of recency when clustering on features

• For example, we can create features for the following:

  • Number of clicks in L12M
  • Number of clicks in L6M
  • Average number of clicks per week

Measuring Aggregates of Event Properties

• There are many different aggregates that can be used for clustering

• The following are a few example:

  • Counts
  • Means
  • Sums

References

• Textbook for Fighting Churn with Data
• Textbook about Algorithmic Marketing
• Analysis about Time-to-Churn using pysurvival
• Course about Predicting Customer Churn
• Course about Customer Churn