The Effect of the Length of the Customer Event History and the Staying Power of the Predictive Models in the Customer Churn Prediction: Case Study of Migros Sanal Market

Abstract

Müşteri terki tahmin problemi çeşitli sektörler için farklı yönlerden incelenmiştir. Bu çalışmada, Türkiye'de önde gelen çevrimiçi hızlı tüketim ürünleri perakendecisinin kayıp tahmin problemi için müşteri olay tarihçesinin uzunluğunun ve öngörücü modellerin kalıcı gücünün etkisini ele alınmaktadır. Bunlar, karar vericilere, müşteri kaybını öngörmek için geçmiş verilerin en uygun uzunluğunu ve öngörücü modellerin ömrünü belirlemede yardımcı olmaları açısından önemli etkilerdir. Çalışma, müşteri olay tarihçesinin uzunluğunun, modellerin kestirimci gücünü  logaritmik olarak artırdığını göstermekte ve literatürde gazete aboneliği sektöründe elde edilen bulguları doğrulamaktadır. Hızlı tüketim sektöründeki müşteri terki modellerin kalıcı gücünün literatürde bir Internet servis sağlayıcısı ve bir sigorta şirketi için belirtilenlerden daha yüksek olduğu görülmektedir.

Keywords

• Müşteri terki
• yapay öğrenme
• gözetimli öğrenme
• gözlem süresi uzunluğu
• modellerin kalıcı gücü

Project Number

3150376

The Effect of the Length of the Customer Event History and the Staying Power of the Predictive Models in the Customer Churn Prediction: Case Study of Migros Sanal Market

Abstract

The customer churn prediction problem is studied for various sectors under several aspects. In this study, we consider the effect of the length of the customer event history and the staying power of the predictive models for the churn prediction problem of a leading online fast-moving consumer goods retailer in Turkey. These are important aspects of the churn prediction models as they help decision makers to determine the optimal length of the past data for predicting the customer churn as well as lifespan of the predictive models. We find that the length of the customer event history logarithmically increases the predictive power of models, validating findings in the literature in the newspaper subscription sector. Regarding the staying power of the predictive models, we conclude that the models in online fast-moving consumer goods retailing has a slightly longer lifespan that the models discussed in the literature for an Internet service provider and an insurance company.

Keywords

• Customer churn
• Machine learning
• supervised learning
• Length of customer event history
• Staying power

Supporting Institution

TÜBİTAK TEYDEB 1501

Project Number

3150376

Thanks

This research was supported by the Scientific and Technological Research Council of Turkey (TÜBİTAK) under TEYDEB 1501 program, under grant numbered 3150376.

References

1. [1] M. Ballings and D. Van den Poel, 2012. Customer event history for churn prediction: How long is long enough? Expert Syst Appl, 39(18), 13517-13522.
2. [2] L. Breiman, 2001. Random forests. Mach Learn, 45(1), 5-32.
3. [3] W. Buckinx and D. Van den Poel, 2005. Customer base analysis: partial defection of behaviourally loyal clients in a non-contractual FMCG retail setting. Eur J Oper Res, 164(1):252-268.
4. [4] D. R. Cox, 1958. The regression analysis of binary sequences (with discussion). J R Stat Soc Series B Stat Methodol, 20, 215-242.
5. [5] J. H. Friedman, 2001. Greedy function approximation: A gradient boosting machine. Ann Stat, 29(5), 1189-1232.
6. [6] J. H. Friedman, 2002. Stochastic gradient boosting. Comput Stat Data Anal, 38(4), 367-378.
7. [7] N. Glady, B. Baesens, and C. Croux, 2009. Modeling churn using customer lifetime value. Eur J Oper Res, 197(1), 402-411.
8. [8] M. Hernant and S. Rosengren, 2017. Now what? Evaluating the sales effects of introducing an online store, J Retail Consum Serv, Volume 39, 305-313.

9. [9] K. Kim, C.-H. Jun, and J. Lee, 2014. Improved churn prediction in telecommunication industry by analyzing a large network. Expert Syst Appl, 41(15), 6575-6584.
10. [10] Y.-H. Lee, C.-P. Wei, T.-H. Cheng, and C.-T. Yang, 2012. Nearest-neighbor-based approach to timeseries classification. Decis Support Syst, 53(1), 207-217.
11. [11] A. Lemmens and C. Croux, 2006. Bagging and boosting classification trees to predict churn. J Mark Res, 43(2), 276-286.
12. [12] A. Martínez, C. Schmuck, S. Pereverzyev, C. Pirker, and M. Haltmeier, 2018. A machine learning framework for customer purchase prediction in the non-contractual setting. Euro J Oper Res, https://doi.org/10.1016/j.ejor.2018.04.034.
13. [13] F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V. Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, and E. Duchesnay, 2011. Scikit-learn: Machine learning in Python. J Mach Learn Res, 12, 2825-2830.
14. [14] H. Risselada, P. C. Verhoef, and T. H. Bijmolt, 2010. Staying power of churn prediction models. J Interact Market, 24(3), 198-208.
15. [15] R. T. Rust and A. J. Zahorik, 1993. Customer satisfaction, customer retention, and market share. J Retailing, 69(2), 193-215.