Lifetime Customer Value
Goal
Provide a score for each existing customer that takes into account purchase history. The score incorporates purchase data that takes into account both historical purchases and forecasted purchases. This amalgamation of historical and future value to the business enables marketing teams to focus resources on customers that have the greatest revenue potential.
Deliverables
- Individual-level LCV labels (i.e., high, medium, low)
- Individual level diagnostic categories (e.g., total spend, recency, average spend, number of transactions, loyalty span, etc.)
Primary Source Data
- Transaction data (minimum with primary key)
- Customer/Contact bio/profile/descriptions
- Transaction data is not necessary once a model has been trained and scored using transaction data. Subsequent LCV scoring can be done with match scores and profile data (e.g., sex, age, zipcode, tel prefix, email domain, etc.)
Required Scores
- Brand/product match scores
Steps
- ETL: date to integer
- record-level transaction arrays
- record-level regression
- extract spend summary values:
- max spend
- total spend
- number of transactions
- extract date summary values:
- earliest transaction date
- most recent transaction date
- span of transaction dates (i.e., number of days)
- number of days between ‘today’ and most recent transaction date
Transformations for customers with <2 transactions
- future spend
- recency
- spend total
- spend maximum
