28
Nov 18

Three Steps to Maximise Customer Retention through the Power of AI and Machine Learning

by Maggie Nazer, Kaloyan Stefanov
Share:

We’re excited to announce that our innovative approach to customer retention through machine learning and predictive analytics has been selected as a finalist in the Innovation and Data Analytics category at this year’s Market Research Society’s (MRS) awards. Our nomination serves as a testimony to our cutting-edge vision and capabilities, and further strengthens our desire to provide our clients with revolutionary solutions to key business challenges. In the words of Kaloyan Stefanov, Managing Director of Data Analytics at GemSeek: “We are committed to pushing the boundaries of innovation in research and analytics, and grateful to our partners for co-creating an exciting machine-learning-based prediction engine.”

Data-driven predictions of customer behaviour are the key to making customer-centric organisations successful. With fierce competition, service providers have to continuously wrestle with increasing churn rates and pressures on customer service. This article sheds light on three innovative and practical steps we undertook in helping a major European Telco overcome the challenge of low customer retention.

Our approach at a glance

Our challenge was to help one of Europe’s leading Telco’s to identify its at-risk customers. To achieve this and help minimise churn we built state-of-the-art machine learning algorithms to predict customer dissatisfaction using several data sets available to the company. We then helped our client embed the predictive model into their customer service operational system, empowering them to proactively approach likely-to-leave customers. The intervention, assisted by a call prioritisation model, enabled our client to reduce the number of customers at risk, achieving a 34 percent uplift in retention.

The path to successful retention in three steps:

  1. Bridge the gap between research and internal data

With most large companies investing in primary research, the fusion of these research findings with existing data stored in large internal databases is a common challenge. However, when applied to the entire customer base of a company, findings from survey research facilitate a better understanding of individual customer perceptions, and become a powerful tool for successful customer targeting, increased sales, and improved retention.  Utilising machine learning to blend and enrich all available customer data sets maximises ROI, and boosts operational synergies throughout teams and departments.

Predicting customer behavior is a cornerstone to proactive customer retention. Yet, gathering personalized customer insights at scale is not only expensive, but often close to impossible, given customers’ resistance to provision of information and their low participation in customer surveys. In a business context, whereby an estimated 10% of clients respond to providers’ queries, customers likely to churn are largely undiscoverable, lest digital footprint data is utilized to the fullest.

 In the case of our Telco client, we developed an AI model that assigned an individual score to every non-respondent of previous client surveys. The model identified dissatisfied customers based on their behaviour, customer service interactions, as well as their product purchases and usage patterns. These predictions enabled our client to target this segment leading to a churn reduction of 38% and our approach was seven times more effective than targeting customers for follow-up calls at random. Our predictions also enabled our client to target customers in real-time and at scale and expanded the targeting of rescue initiatives from ca. 10% of the addressable customers to all addressable customers.

  1. Figure out what truly matters

Different characteristics of products, services, and customer experience influence customers’ decisions to remain loyal to varying degrees. Knowing what is important to customers not only requires constant development, it is also paramount when taking measures to counteract churn. As we helped our Telco partner get better aligned with its customers, we used a gradient boosting model to calculate the impact of over 1,200 variables. Apart from being able to prioritise predictors of either loyalty or churn, we also discovered important dependencies between variables and their respective impact on the likelihood of churn.

  1. Revamp your company-wide retention process

Whereas machine learning allowed us to identify the most important factors for churn and predict customer behaviour, the business value of cutting-edge customer retention projects based on machine learning becomes apparent once the resulting models are fully integrated into customer service operations. Automation plays a key role in rendering machine learning insights actionable.

In this instance we were able to embed our model into their CX platform to provide details of those customers at risk of leaving in real time. Lists were generated on a daily basis, and were based on scripts trained to autonomously run the entire machine learning process every month.

Naturally, customer service agents become more productive and effective when given clear instructions on which specific customers to target, and how to turn them into promoters by offering what truly counts. In the case of our client, our machine learning model served as the inspiration for a complete revamp in customer service operations. In addition to significantly reducing churn rates, the company adopted new strategies to record and analyse retention efforts to maximise effectiveness.

 

A smart investment

            Depending on your industry, retaining an existing customer can be up to 30 times less costly than acquiring a new one. In times of vast competition and high customer expectations, customer retention should be tackled with utmost precision and foresight. Whereas we have already argued against applying machine learning to just any type of business question at hand, reducing customer churn with the help of machine learning has significant financial benefits, and one with unparallel returns. The case study we presented in this article is a great illustration of our ground-breaking work to harness the power of machine learning for long-term customer retention.

*Did we spark your interest? GemSeek is a hub of business-savvy data scientists. We pair superb data knowledge and data science skills with client-centric consultancy to find the best solutions to each of our partner’s challenges and turn it into an opportunity. E-mail kaloyan.stefanov@gemseek.com to get in touch, and chat about the best ways to bring machine learning into your customer experience pipeline. Alternatively follow our Facebook page for fresh news and updates.