1 |
en bedrijfskunde
25-04-2018
Predicting churn based on the interplay
between usage (dynamics) and newsletter
email marketing
Thesis defense
MSc. Marketing Intelligence and Marketing Management
28 January 2019
Dick van Klaarbergen
S2749351
› Customer retention management has a large impact on firm
performance
Higher than customer acquisition
(Rust & Zahorik, 1996; Pfeifer, 2005)› Churn prediction to increase customer retention management
effectiveness
Churn: customer that uses another learning method than the year
before
- Contractual churn
› In B2B churn is different
Little churn literature in B2B
Balance importance satisfaction and switching costs is different
(Palmatier et al., 2006; Jones et al., 2000) - Influences churn prediction
Retention more effective in B2B
(Rauyruen & Miller, 2007; Narayandas,2005; Mudambi, 2002)
Predicting churn is useful
The relation of usage and newsletter emails
on churn is studied
How is the interplay between usage (dynamics) and
newsletter email marketing related to the churn probability
of a customer?
› Educational publisher data
Core courses secondary education
- B2B
Analysis level
: school-section-course
› Data
Churn data: based on booklist data
Usage data: based on scores on online learning method
- Only online part of usage
Newsletter email data: based on newsletters sent
› 11,816 observations
466 (3.94%) churners
Educational publisher data is used
› Binary logistic regression model
DV: churn
› Model performance criteria
AIC: parsimonious model
TDL: predictive performance
› Advanced machine learning techniques
(by using caret)
Decision trees
C5.0
Neural Networks
Ensemble learning
(bagging, boosting & random forest)- Balanced sampled machine learning
Logistic regression is used to model churn
Introduction Theory Data Method
Results
Discussion
Significant predictors
› RFI usage
Recency Churn (+)
Frequency Churn (-)
Frequency moderates
Recency Churn (+)
(see figure)
› Dynamics in usage
Monthly dips moderates Frequency Churn (+)
Monthly variation churn (-)
› Newsletter marketing
Newsletters Churn (-)
› Targeting customers by
using usage data for
newsletter emails is not
effective in reducing churn
› Effective to predict churn
with usage data
TDL of 2.3 (see figure)
› Logit performs better than
advanced machine learning
techniques
Balanced sampling causes
instable performance
The model performs more than 2 times
better than a naïve model
› Theoretical implications
Effect of intensity and most dynamic factors not
approved
Effect of targeting email newsletters by usage data not
approved
Peaks indicate something else than dips
Machine learning not effective
› Managerial implications
Use churn predictions for targeted emails
Use churn prediction for prioritizing of retention efforts
-
Like account manager visits
The model can be used for targeting in
retention efforts
10 |
en bedrijfskunde
25-04-2018
References
› Jones, M. A., Mothersbaugh, D. L., & Beatty, S. E. (2000). Switching
barriers and repurchase intentions in services. Journal of Retailing, 76(2), 259-274.
› Narayandas, D. (2005). Building loyalty in business markets. Harvard
Business Review, 83(9), 131-139.
› Palmatier, R. W., Dant, R. P., Grewal, D., & Evans, K. R. (2006). Factors influencing the effectiveness of relationship marketing: A meta-analysis.
Journal of Marketing, 70(4), 136-153.
› Pfeifer, P. E. (2005). The optimal ratio of acquisition and retention costs.
Journal of Targeting, Measurement and Analysis for Marketing, 13(2),
179-188.
› Rauyruen, P., & Miller, K. E. (2007). Relationship quality as a predictor of B2B customer loyalty. Journal of Business Research, 60(1), 21-31.