Tiered Modeling

Defining Tiered Modeling for Segmentation

• Typically, a targeting model scores the relevance of a customer in terms of a business objective based on their customer profile

• One of the most basic approaches is to score customers based on one metric alone

  • E.g. average spend of a customer
• This feature can then be used to sort customers by a metric into tiers to identify the most valuable and least valuable customers

• However, it's common to create tiers based on multiple metrics as well

  • E.g. loyalty-monetary segmentation

• A standard example is to assign each customer to one of 5 tiers:

  • The top 5% of customers to tier 1
  • The next top 10% to tier 2
  • The next top 20% to tier 3
  • The next top 30% to tier 4
  • The bottom 35% to tier 5
• Then, we can profile each tier to understand customers' average monetary and behavioral attributes

Illustrating Manual Steps for Tiered Modeling

1. Assign quantile scores to each customer

   • These scores represent quantiles (e.g. $1-5$) based on some metric (e.g. spend)
2. Optional: Aggregate quantile scores
3. Analytically assign scores to segments
4. Profile segments

Illustrating Clustering Steps for Tiered Modeling

1. Assign quantile scores to each customer

   • These scores represent quantiles (e.g. $1-5$) based on some metric (e.g. spend)
2. Optional: Aggregate quantile scores
3. Cluster scores to retrieve segments
4. Profile segments

References

• Textbook about Algorithmic Marketing
• Example of Tiered Modeling
• Course about Customer Segmentation
• Course about Segmentation Analysis