Analyzing Customer Cohorts

Sample analysis of customer cohort for improved marketing targeting. By identify the most valuable customer segments, audience targeting can be improved along with the creation of look a like audiences.

## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

Data Wrangling

First, remove negative values, null values, and incomplete months

## [1] 541909      8

## [1] 531285      8

## [1] 397924      8

## [1] "Date Range: 2010-12-01 08:26:00 ~ 2011-12-09 12:50:00"

## [1] 397924      8

## [1] 380620      8

Data Normilzation

##    CustomerID       TotalSales.V1        OrderCount.V1       AvgOrderValue.V1  
##  Min.   :12346   Min.   :-1.7314464   Min.   :-1.7314464   Min.   :-1.7314464  
##  1st Qu.:13815   1st Qu.:-0.8660254   1st Qu.:-0.8657232   1st Qu.:-0.8657232  
##  Median :15300   Median : 0.0000000   Median : 0.0000000   Median : 0.0000000  
##  Mean   :15302   Mean   : 0.0000000   Mean   : 0.0000000   Mean   : 0.0000000  
##  3rd Qu.:16781   3rd Qu.: 0.8657232   3rd Qu.: 0.8657232   3rd Qu.: 0.8657232  
##  Max.   :18287   Max.   : 1.7314464   Max.   : 1.7314464   Max.   : 1.7314464

##    CustomerID    TotalSales    OrderCount AvgOrderValue 
##      1720.983         1.000         1.000         1.000

Customer Cluster Visualizations

##   TotalSales OrderCount AvgOrderValue
## 1  1.2058544  1.0068055     0.8683349
## 2 -0.1314289 -0.8520880     0.7984689
## 3  0.2234497  0.7188893    -0.6402441
## 4 -1.2412204 -0.7892257    -1.0603240

## # A tibble: 4 × 2
##   Cluster Count
##     <int> <int>
## 1       1  1134
## 2       2  1062
## 3       3   943
## 4       4  1159

ggplot(normalizedDF, aes(x=TotalSales, y=OrderCount, color=Cluster)) +
  geom_point()

ggplot(normalizedDF, aes(x=TotalSales, y=AvgOrderValue, color=Cluster)) +
  geom_point()

Cluster Silhouette

## [1] "Silhouette Score for 4 Clusters: 0.4120"
## [1] "Silhouette Score for 5 Clusters: 0.3761"
## [1] "Silhouette Score for 6 Clusters: 0.3773"
## [1] "Silhouette Score for 7 Clusters: 0.3915"
## [1] "Silhouette Score for 8 Clusters: 0.3808"

# Interpreting customer segments
cluster <- kmeans(normalizedDF[c("TotalSales", "OrderCount", "AvgOrderValue")], 4)
normalizedDF$Cluster <- cluster$cluster
# count per cluster
normalizedDF %>% group_by(Cluster) %>% summarise(Count=n())

## # A tibble: 4 × 2
##   Cluster Count
##     <int> <int>
## 1       1   950
## 2       2  1136
## 3       3  1150
## 4       4  1062

# cluster centers
cluster$centers

##   TotalSales OrderCount AvgOrderValue
## 1  0.2132451  0.7112607    -0.6432176
## 2  1.2059015  1.0076634     0.8661868
## 3 -1.2460083 -0.7960747    -1.0616569
## 4 -0.1314289 -0.8520880     0.7984689