Welcome to the co u rse MAR K E TIN G AN ALYTIC S : P R E D IC TIN G - PowerPoint PPT Presentation

Welcome to the co u rse MAR K E TIN G AN ALYTIC S : P R E D IC TIN G C U STOME R C H U R N IN P YTH ON Mark Peterson Senior Data Scientist , Alliance Data

Ch u rn Anal y tics MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

C u stomer ch u rn When an e x isting c u stomer stops doing b u siness w ith a compan y MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Contract u al ch u rn MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Vol u ntar y ch u rn MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Non - contract u al ch u rn MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

In v ol u ntar y ch u rn : Credit card e x piration MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

In v ol u ntar y ch u rn : Utilities t u rned off MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Utili z ing y o u r e x perience C u stomer Lack of u sage Poor Ser v ice Be � er Price Domain / ind u str y kno w ledge MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Telco Ch u rn Dataset Description Val u e Records 3333 Feat u res 21 Contino u s 15 Categorical 6 MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Feat u res of interest Voice mail International calling Cost for the ser v ice C u stomer u sage C u stomer ch u rn MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Ho w ch u rn is defined here C u stomer cancelling their cell u lar plan at a gi v en point in time "no" "yes" MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

E x plorator y data anal y sis u sing pandas Understand the feat u res of the dataset Comp u te s u mmar y statistics MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

E x plorator y data anal y sis u sing pandas pandas Fo u ndations df.head() df.describe() df.mean() MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Let ' s e x plore the data ! MAR K E TIN G AN ALYTIC S : P R E D IC TIN G C U STOME R C H U R N IN P YTH ON

Gro u ping and s u mmari z ing data MAR K E TIN G AN ALYTIC S : P R E D IC TIN G C U STOME R C H U R N IN P YTH ON Mark Peterson Senior Data Scientist , Alliance Data

Ch u rners and non - ch u rners print(telco['Churn'].value_counts()) no 2850 yes 483 Name: Churn, dtype: int64 MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Model o u tcomes T w o classes : 'yes' : C u stomer w ill ch u rn 'no' : C u stomer w ill not ch u rn MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Differences bet w een ch u rners and non - ch u rners Do ch u rners call c u stomer ser v ice more o � en ? Does one state ha v e more ch u rners compared to another ? MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Gro u ping and s u mmari z ing data .groupby() MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Let ' s gro u p and s u mmari z e ! MAR K E TIN G AN ALYTIC S : P R E D IC TIN G C U STOME R C H U R N IN P YTH ON

E x ploring y o u r data u sing v is u ali z ations MAR K E TIN G AN ALYTIC S : P R E D IC TIN G C U STOME R C H U R N IN P YTH ON Mark Peterson Senior Data Scientist , Alliance Data

Vis u ali z ing data in P y thon seaborn librar y allo w s y o u to easil y create informati v e and a � racti v e plots B u ilds on top of matplotlib MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Vis u ali z ing the distrib u tion of acco u nt lengths Important to u nderstand ho w y o u r v ariables are distrib u ted import matplotlib.pyplot as plt import seaborn as sns sns.distplot(telco['Account_Length']) plt.show() MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Differences in acco u nt length Bo x plot sns.boxplot(x = 'Churn', y = 'Account_Length', data = telco) plt.show() MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Differences in acco u nt lengths Bo x plot sns.boxplot(x = 'Churn', y = 'Account_Length', data = telco) plt.show() MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Differences in acco u nt lengths Bo x plot sns.boxplot(x = 'Churn', y = 'Account_Length', data = telco) plt.show() MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Differences in acco u nt lengths Bo x plot sns.boxplot(x = 'Churn', y = 'Account_Length', data = telco) plt.show() MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Differences in acco u nt lengths Bo x plot sns.boxplot(x = 'Churn', y = 'Account_Length', data = telco) plt.show() MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Differences in acco u nt lengths Bo x plot sns.boxplot(x = 'Churn', y = 'Account_Length', data = telco) plt.show() MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Differences in acco u nt length Bo x plot sns.boxplot(x = 'Churn', y = 'Account_Length', data = telco) plt.show() MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Differences in acco u nt length Bo x plot sns.boxplot(x = 'Churn', y = 'Account_Length', data = telco, sym="") plt.show() MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Adding a third v ariable sns.boxplot(x = 'Churn', y = 'Account_Length', data = telco, hue = 'Intl_Plan') plt.show() MARKETING ANALYTICS : PREDICTING CUSTOMER CHURN IN PYTHON

Let ' s make some plots ! MAR K E TIN G AN ALYTIC S : P R E D IC TIN G C U STOME R C H U R N IN P YTH ON