Overview of Segmentation

Defining Customer Segmentation

• Generally, there are two types of customer segmentation:

  • Behavioral segmentation
  • RFM analysis
• RFM analysis involves segmenting customers based on the observed financial results
• Behavioral segmentation involves identifying the traits causing these results
• Specifically, behavioral segmentation clusters customers based on

Describing RFM Analysis

• RFM analysis is based on three different metrics:

  • Recency
  • Frequency
  • Monetary

• The recency metric represents the number of time units having passed since the customer last purchased

  • This metric can be measured directly in:

    • Time units (e.g. months)
    • Or some score
  • Customers can be sorted by the most recent purchase date
  • Then, those customers in the most recent $20 \%$ are assigned a score of $5$
  • Maybe, the next $20 \%$ have a score of $4$
  • And so on, until the last $20 \%$ get a score of $1$

• The frequency metric represents the average number of purchases per time unit

  • Again, the metric can be measured directly in units or scores

• The monetary metric represents the dollar amount spent per time unit

  • It is typically measured using intervals or scores

• It is quite typical to use the same discrete scoring scale (e.g. $1-5$)

  • The same scale can be used for all $3$ metrics
• Recency, frequency, and monetary metrics are often correlated with the probability to respond and the LTV

• RFM approach is shallow

  • It only measures the final outcomes of the marketing processes and consumer actions
  • Not, the factors that impact consumer behavior
  • A more flexible solution may be behavioral clustering

Describing Behavioral Segmentation

• Behavioral segmentation defines a small number of clusters with semantic meaning

• The output of the segmentation process typically includes:

  • Segment profiles
  • Segment models

• A segment profile includes:

  • The distinctive properties and metrics of the segment
  • Some interpretation of what a typical cluster looks like

• These distinctive properties are identified using clustering algorithms

  • Specifically, by running clustering algorithms on a set of historical customer profiles
  • So, each segment corresponds to a group of existing customers
  • The segment profile is a set of statistical metrics for this group
• The following table consists of customer profiles:

Persona	Convenience Seekers	Casual Buyers	Bargain Hunters
% of Market	20	50	30
% of Revenue	40	40	20
Share of Clothing	40	60	60
Share of Electronics	50	20	10
Share of Toys	10	20	30
Redemption Rate	0.02	0.05	0.08

• Note, the convenience seekers, casual buyers, and bargain hunters columns were all learned through clustering algorithms
• Meaning, Segment $1$ represents convenience seekers
• Segment $2$ represents casual buyers
• Segment $3$ represents bargain hunters

Using Customer Segmentation in Targeting Systems

• Behavioral segmentation can be used for creating new behavioral features about customers

• Often, these customer segments can be used as predictors in:

  • A look-alike model
  • Other targeting models
• These segments carry an important signal about consumer behavior
• As a result, they can have substantial predictive power for propensity modeling
• However, propensity models created for an entire population of customers can have limited accuracy
• This is because propensities can be determined by different factors

• For example, customers in one segment can churn because of low product quality

  • But, customers in another segment churn because of high prices

• As a result, we should usually have an individual model for different:

  • Business objectives
  • Product categories
  • Customer segments

References

• Textbook about Algorithmic Marketing